Towards the construction of a validated numerical system to study the mesoscale dynamics of the North East Atlantic (2003-2006)

UNIVERSIDAD DE LAS PALMAS DE GRAN CANARIA Facultad de Ciencias del Mar Departamento de Física Proyecto Final de Máster – Tesina: Towards the construction of a validated numerical system to study the mesoscale dynamics of the North-East Atlantic (2003-2006) case study. Sheila Natalí Estrada-Allis Director: Ángel Rodríguez Santana Co-directores: Rui Caldeira Andrade y Pablo Sangrà Inciarte Las Palmas de Gran Canaria a 3 de Febrero de 2011 Towards the construction of a validated numerical system to study the mesoscale dynamics of the North East Atlantic (2003-2006) S.N. Estrada-Allisa,c, X. Couvelardb, R.M. Caldeirab,c,1,∗, F. Mach´ına, A. Rodr´ıguez-Santanaa, P. Sangr`aa aDepartamento de F´ısica. Facultad de Ciencias del Mar. Universidad de Las Palmas de Gran Canaria (ULPGC). Las Palmas. Spain bCenter for Mathematical Sciences (CCM) University of Madeira. Madeira Island. Portugal cInterdisciplinary Centre of Marine and Environmental Research (CIIMAR). Porto. Portugal Abstract A detailed validation study has contributed to the construction of a NEAtlantic (NEA), ocean circulation model, for the 2003-2006 period. The comparisons between three model solutions, remote sensing and in situ data, focused on the study of the most dynamical processes of NEA sub-regions. Model validations include (i) comparisons with Sea Surface Temperature (SST), from AVHRR and Microwave-OI; (ii) Eddy Kinetic Energy (EKE), computed from altimetry; (iii) Temperature and Salinity profiles computed from ARGO floats, and (iv) sub-surface temperature, measured in three buoys of ‘Puertos del Estado’. Simple statistical methods were used to quantify model-data comparisons. In general, model regional solutions show a ∗Corresponding author Email address: [email protected] (R.M. Caldeira) 1CIIMAR, Rua dos Bragas, 289, 4050-123 Porto, Portugal, Ph (+351) 22 340 18 00, Fax (+351) 22 339 06 08 Preprint submitted to Ocean Modelling August 10, 2011 good correlation and small Root Mean Squared Error (RMSE) with SST and EKE. The main water masses were well depicted by the regional model; while the region with high salinity values, often dominated by Mediterranean Intermediate Water (MIW), was not accurately resolved. The initial condition and boundary forcing of the regional model was evaluated, particularly concerning the usage of Ocean General Circulation Models (OGCM’s), as an alternative to the classical climatological forcing. The analysis of their Kolmogorov energy spectrum determined their effective resolutions and their EKE levels. Mercator global solution at 1/4◦, was shown to be an adequate OGCM solution for studying the 2003-2006 period. The energy spectrum analysis also showed that new 1/12◦(downscaled), regional solution resolved more energetic scales than the original OGCM, confirming the need to use high spatial resolution regional ocean circulation models to resolve mesoscale and sub-mesoscale phenomena. The regional model (ROMS) was able to reproduce an Iberian Peninsula upwelling event as well as other previously documented NEA processes, such as the westward propagating eddies associated with the Azores Front, with the Canary and with Madeira Archipelagos. The analysis also showed that atmospheric forcing, is important to adequately resolve surface dynamics. However, other more conservative aspects of the ocean such as the water mass composition, are better represented with climatological forcing. Keywords: ROMS, North Eastern Atlantic, model validation, atmospheric-ocean interactions, boundary conditions. 2 1. Introduction The North East Atlantic ocean (NEA) is a very dynamic oceanic region with identifiable sub-systems; these sub-systems can be considered as “critical regions”, characterized by the high values of eddy kinetic energy (EKE), (Fig. 1). The main critical regions of the NEA, identified using al-5 timetry derived data for their high EKE activity (Fig. 1), are: (i).- Iberian Peninsula (IP); (ii).- The Mediterranean Water Outflow (MWO); (iii).- The Azores Front region (AF); (iv).- Madeira Archipelago (MA); (v).- Canaries Archipelago (CA) and (vi).- The Azores Archipelago (AA). The IP sub-region is part of the four main driven eastern boundary up-10 welling zones, hereafter refer to as IPUS (Iberian Peninsula Upwelling System), with a strong seasonal variability (e.g. Alvarez et al., 2009). Two water masses play an important role: (i) the Mediterranean Intermediate Water (MIW), since the upwelling is in the pathway of anticyclonic mesoscale lens of warm salty of Mediterranean Water (Meddies) (e.g. Armi and Stommel,15 1983; Relvas et al., 2007); (ii) the Eastern North Atlantic Central Water (ENACW), that is, in general, observed in upwelled waters (e.g. Pollard et al., 1996). In absence of coastal upwelling, the surface circulation off Western Iberia is predominantly poleward (e.g. Peliz et al., 2002). Another important and recurrent structure in the IPUS are the upwelling filaments20 (e.g. Barton et al., 2001), as well as the long trapped filaments (see Meunier et al., 2010). Satellite data (Haynes et al., 1993) and models results (Haidvogel et al., 1991) show that large filaments are often closely related with the location of capes and promontories. Laboratory studies by Mauritzen et al. (2001) hypothesize that the MWO25 3 Mediterranean Water Outflow exerts a strong influence in the North East Atlantic dynamics. Downstream of the Gibraltar Strait, higher salinities were measured in the surface waters of the Gulf of Cadiz, through a detrainment process (i.e. due to diapycnal mixing of flux of salinity from the MWO towards the low-density Central Water). The surface waters travel westward30 and northward, promoting the salinity increase in the Central Water off the West IP (see Relvas et al., 2007). Furthermore, previous modeling studies suggest the existence of a direct link between the Azores Current and the MWO (e.g. Relvas et al., 2007; Volkov and Fu, 2010). The formation of the well-defined zonally oriented Azores Current may be the result of water mass35 transformation associated with MWO in the Gulf of Cadiz (Volkov and Fu, 2010). The Azores Current, and its associated front (AF), is a quasi-permanent NEA feature throughout the year, centered between 33◦and 35 ◦N (e.g. Klein and Siedler, 1989). The complex mesoscale variability of the AF is char-40 acterized by long periods (∼250 days) and large wavelengths (∼600km), with westward propagation, associated with Rossby waves (Le Traon and De Mey, 1994) and are largely due to baroclinic instabilities (e.g. Alves and De Verdi`ere, 1999). Chelton et al. (2007), also suggested that the westward energy propagation at mid-latitudes is more representative of nonlinear45 vertically-coherent eddies, as an alternative to the Rossby wave propagation theory. Islands are also regions with strong mesoscale activity in the NEA. Sangr`a et al. (2009) reported the existence of several west propagating eddy corridors at the AF and leeward of the islands. Two small eddies corridors were50 4 identified north and south of the AF. These west propagating cyclonic eddies were first observed at 32.2 ◦N from in-situ data (Pingree and Sinha, 2001) and presented as an alternative hypothesis, which had previously associated these westward propagation with the occurrence of planetary Rossby waves. Another zonal corridor was also detected at 31 ◦N, south of MA and south55 of the CA. The Canary Eddy Corridor, extending from 22 ◦N to 29 ◦N, and populated mainly by anticyclones. Both Madeira and Canary eddy corridors, had well defined source regions i.e. leeward side of the islands. The first attempts to study the dynamical mesoscale sub-systems of the NEA, using only Ocean Global Circulation Models (OGCM), did not repro-60 duce well the mesoscale structures often observed from in-situ and remote sensing data (e.g Pingree, 2002). Mesoscale eddies are parameterized as a viscous term playing an important role in energy and momentum dissipation in ocean circulation models, not well resolved in OGCMs (Jochum et al., 2008). Another OGCM limitation is their direct impact on deep-ocean65 currents, which are largely constrained by the poor representation of the oceanic bathymetry (Barth et al., 2008). Nevertheless, OGCMs can provide adequate boundary and initial conditions for high-resolution regional ocean models (Barth et al., 2008; Melsom et al., 2009; Alvera-Azc´arate et al., 2011). Currently, there are several OGCM solutions freely available to the scientific70 community. These include: (i) Mercator, (ii) SODA, (iii) ECCO and (iv) HYCOM. In order to determine which OGCM to use, their accuracy and availability of their solutions were considered herein. Classical climatological forcing was also considered, as an alternative control experiment. Previous studies (Mason et al., 2011) used high resolution regional models75 5 to study some aspects of the NEA dynamics, forcing their model with climatological data. To the best of our knowledge, no previous studies evaluated the relative role of OGCMs and climatological boundary forcing, to study the NEA dynamics. Ultimately, our main goal is to construct a validated ocean regional model, considering the tools and the data available today, in order80 to adequately continue to investigate the regional dynamics. It is expected that this system and its validation protocols will evolve into a regional ocean forecasting system. The layout of this report is as follows, after the introduction (section 1), Section 2 describes the different sources of data used, as well as, a detailed85 description of the regional numerical modeling system, the different aspects of the experiments, their boundary and initial conditions and the statistical metrics used for comparing model results with data. Section 3 discusses the main results including: (i) The representation of sub-mesoscale NEA processes; (ii) The representation of the surface dynamics; and (iii) the ca-90 pability of our regional model to reproduce previously documented dynamic processes, such as the westward propagation eddies, the representation of main water masses and an IPUS episode. Section 4 sums up the results and proposes future directions. 2. Data sources and Methods95 2.1. The regional ocean circulation modeling system The regional model used in this study was the Regional Oceanic Modeling System (ROMS). For a complete description of the model referred to (Shchepetkin and McWilliams, 2003, 2005). ROMS is a split-explicit, free6 surface and terrain-following vertical coordinate oceanic model, where short100 time steps are used to advance the surface elevation and barotropic momentum equation, and a larger time step is used for temperature, salinity, and baroclinic momentum. ROMS employs a two-way time-averaging procedure for the barotropic mode which satisfies the 3D continuity equation. The specially designed predictor-corrector time-step algorithm allows a substantial105 increase in the permissible time-step size. The third-order, upstream-biased, dissipative advection scheme for momentum allows the generation of steep gradients, enhancing the effective resolution of the solution for a given grid size Shchepetkin and Mcwilliams (1998). For tracers, the RSUP3 scheme where diffusion is split from advection and is represented by a rotated bi-110 harmonic diffusion scheme with flow-dependent hyper-diffusivity, is used in order to avoid excessive spurious diapycnal mixing associated with sigma coordinates Marchesiello et al. (2009). Explicit lateral viscosity is null everywhere in the model, except in sponge layers near the open boundaries where it increases smoothly on several grid points. A K-profile parameterization115 (KPP) boundary layer scheme Large et al. (1994) parameterizes the sub-grid vertical mixing processes. In order to encompass the most relevant dynamic features of the NEA circulation, and considering the oceanographic dataset available for model validation, we have designed an extended rectangular grid from 25 ◦N to 45120 ◦N in latitude and from 35 ◦W to 5 ◦W in longitude. The model grid, forcing initial and boundary conditions are built using an adapted version of the ROMSTOOLS package (Penven, 2003). The bottom topography is derived from a 30 arc-second resolution database GEBCO 08 7 (www.gebco.net). Although a new pressure gradient scheme associated to a125 modified equation of state limits computational errors of the pressure gradient is currently implemented in ROMS (Shchepetkin and McWilliams, 2003), the bathymetry still needs to be smoothed, so that the “slope parameter” r=∆h/h (Beckmann and Haidvogel, 1993) remains under 0.2. To preserve a sufficient resolution in the upper ocean, we use 50 vertical levels130 with stretched s-coordinates, using surface and bottom stretching parameters θs=6, θb=0 (Song and Haidvogel, 1994). Three ROMS experiments were built for this study, they differed on the source and nature of their oceanic boundary conditions and on the source and nature of their atmospheric forcing conditions. These include: (1) R M,135 ROMS forced with Mercator at the oceanic boundaries. Half degree, daily mean wind stress was extracted from the QuikSCAT satellite scatterometer data, provided by CERSAT (www.ifremer.fr/cersat/en/index.htm); heat and fresh water fluxes were extracted from NCEP2, using the bulk formula Fairall et al. (1996, 2003); (2) R NQ, ROMS was forced with WOA05 climatology140 (Locarnini et al., 2006; Antonov et al., 2006) at the ocean boundary, momentum fluxes were extracted from QuiKSCAT and heat fluxes from NCEP2; (3) R C, ROMS forced with WOA05 climatology at the oceanic boundaries, and COADS climatology (da Silva et al., 1994) (atmospheric heat and momentum).145 The lateral boundaries facing the open ocean, a mixed passive-active, implicit, radiation condition connects the model solution to the surroundings (Marchesiello et al., 2001). Regarding the Mercator inflow conditions, the solution at the boundary is nudged toward daily time-averaged outputs, which 8 (sub)mesoscale variability, there is a need to use high-resolution model grids. Comparing the OGCM solutions (figure 4), Mercator (1/4◦) and HYCOM (not shown) resolve well up to the 100 km range, but it needs to be coupled with ROMS, in order to fully resolve the sub-mesoscale variability (10-100 km). On the other hand, ECCO and SODA are only representing larger295 scales (100-1000 km). Although this analysis is done for the same region, and for a particular (random) moment in time, it is not expected that it will vary much over space and time. Therefore, the ability of numerical ocean models to adequately resolve the energy spectrum for the different scales is very much a function its grid resolution. The enstrophy transfer which is300 diagnosed by k−3, is only resolved by the higher resolution regional models i.e. ROMS. There are also noticeable differences between the spectral density resolved by climatological forced regional model (R C) and the other ROMS solu-305 tions (e.g. R M;R NQ). Climatological forced ROMS resolve processes with lower spectral density signatures (10−2versus10−1revolved by the other ROMS experiments, R M;R NQ). Furthermore, the atmospherically forced ROMS show a greater turbulent variability, when compared with the climatologically forced experiment.310 3.2. Regional ocean circulation model validation The validation of our regional ocean circulation model was achieved studying some of the previously documented dynamics for the NEA critical subregions. First, the surface dynamics were studied by comparing model solutions with EKE, SST and buoys data. Secondly, the west propagation of315 15 mesoscale eddies were studied comparing model to AVISO data. Thirdly, attention was given to the model reproducibility of the main NEA water masses. Finally, a case study of an Iberian Peninsula Upwelling episode, which occurred between the 5th and the 20th August of 2005, helped demonstrate the accuracy of the regional model solution to represent a specific320 oceanic event. 3.2.1. Representation of surface dynamics Barth et al. (e.g. 2008), showed that OGCM boundary conditions improved the regional model solution, particularly on the shelf dynamics, even when the open boundary is located far, in the open ocean. Climatological325 forcing produced low-density gradients and less energetic regional solutions (Barth et al., 2008). Our results show that to adequately reproduce the interannual SST variability (see figures 5) another important factor to consider is the use of appropriate atmospheric forcing, such as better representative winds and heat fluxes. Likewise, in our study better SST comparisons were330 achieved using OGCM forced ROMS than climatological forced 17(Table 4). Experiments R M and R NQ show better comparisons with AVHRR derived SST data in relation to the purely climatologically forced experiment (R C), (Fig. 5). The bias values are closer to 0 n R M and R NQ experiments, while the R C experiment shows a greater variability. The RMSE values vary be-335 tween 0 and 1 in R M and R NQ whereas in R C it reaches 1.9. The analysis of the r2coefficient shows lowest correlation in winter months, probably due to limited data availability i.e. higher cloud coverage (e.g. Reynolds et al., 2007). Non-spite the fact that the satellite SST data uses OI to fill the missing values, it is expected the use of more SST valid data points during cloud340 16 free periods, thus interpolated products are expected to have more accurate representation. The comparisons between AVHRR derived SST and R C worsens with time, 2003 RMSE and Bias are higher for 2006. Correlations are also weaker for 2006 than for the 2003-2005 period. Recent values of the DJFM NAO345 index, shows a progressive change for the same period (2003-2006). NAO indexes are -0.20 in the winter months of 2003, -0.11 in the winter of 2004, -0.82 in the winter of 2005, becoming 1.83 (positive) in the winter of 2006 (http://www.cru.uea.ac.uk/ timo/datapages/naoi.htm). In order to account for the inter-annual atmospheric variability, characteristic of periods with350 positive and negative NAO-years, it is important to consider non-climatological atmospheric forcing conditions. The leading mode of sea surface temperature (SST) variability over the North Atlantic during the positive NAO winter consists of a tripole pattern: (i) with a cold SST anomaly in the sub-polar region, (ii) a warm anomaly in the middle latitudes centered off Cape Hat-355 teras (NEA case); (iii) a cold subtropical anomaly between the equator and 30 ◦N(Deser and Blackmon, 1993; Kushnir, 1994). The temporal averaged SST shows the best model-data comparisons in the open-ocean (Fig. 6), thus there is a need to complement this analysis comparing model results with near-surface temperature collected using in situ360 buoys (see e.g. Barth et al., 2008; Ivanov et al., 2009). Variability of coastal (near-shore) processes are not well represented in merged/interpolated satellite products. The sub-surface (3m) temperature time-series of model solutions are in good agreement with the observations collected by the VS (Villano Sisargas), with the CS (Cabo Silleiro) and EB (Estaca de Bares)365 17 buoys as shown in temporal series of temperature in figure 8. The ROMS experiments using high resolution atmospheric forcing (R M and R NQ) are able to reproduce the inter-annual variability in the buoys, whereas the climatologically forced experiment (R C), showed always weaker correlations. With respect to EKE derived from altimetry data, the best model rep-370 resentation is obtained in offshore regions and using climatological forcing conditions (figure 7). It is hypothesized that these differences are due to (i) the limited spatial and temporal resolution of AVISO data (1/3◦), and (ii) due to the fact that the models have higher resolution compared to AVISO data, and thus they are more energetic. Furthermore, the lack of AVISO data375 near the coast favors the comparisons with offshore regions, since AVISO data accuracy is limited to acquire data up to 45 km from the coast (Durand et al., 2008). 3.2.2. West propagation of eddies Figure 9, shows the westward propagation of mesoscale eddy structures380 (see white arrow in figure 9a and b), (34.0423 ◦N between 19 ◦to 35 ◦W). Sangr`a et al. (2009) observed two small corridors of westward propagating eddies, north and south of Azores Front. These corroborate the results showed in figures 9a and b. Also apparent, in our results, is the fact that these structures live less than 28 weeks (∼200 days). Pingree and Sinha (2001),385 after analyzing infrared, altimeter and in-situ measurements, suggested that the westward movement of these large structures are due to cold (cyclonic) structures called STORMS, propagating westward around 32.2 ◦N. Thereafter, Pingree (2002) showed westward propagating anomalies (in their figure 3), occurring in the same area reproduced by our ROMS experiments390 18 (figure 9c and d). Furthermore, there is good qualitative agreement between AVISO EKE data and the simulated EKE extracted from the R M experiment. Moreover, the mesoscale structures observed in the figure 9c and d, correspond to the seasonal variability of EKE showed in Sangr`a et al. (2009). The highest values of EKE are found south of the CA archipelago in spring,395 summer and autumn. 3.2.3. Water masses reproducibility The waters masses described for the NEA are better depicted by the climatological forced regional model (R C), and represented to a lesser degree of realism in the experiments using high resolution atmospheric forcing. It400 is expected that climatological forced ROMS best reproduces water masses composition due to its conservative nature. Deep-water masses with slow overturning rates, often takes hundreds of years for a complete re-circulation. On the other hand, this also challenges the classical view that atmospheric forcing only affects the ocean surface dynamics. Somehow, through vertical405 mixing (advection / diffusion), atmospheric forcing seems to also play an important role on water mass composition. In fact, there are recent examples in the literature suggesting that mesoscale surface-induced features, have a significant influence in ‘venting’ deep-water masses, perturbing its conservative flow regime.410 The central water, North Atlantic Central Water (NACW; Fig. 10), characterized by density values of 27.38 (kg/m3), with potential temperature varying (θ) between 11-18 ◦C and salinities varying between 35.5-36.5 (Mach´ın et al., 2006), showed good reproducibility by the model. However, the less reproducible water mass was indeed, the warmer and saltier Mediter-415 19 ranean Intermediate Water (MIW; Fig.10). MIW is characterized by θ= 10 ◦C, with S higher than 35.6, and densities varying between 27.38-27.922 (kg/m3) (Mach´ın et al., 2006). The Antarctic Intermediate Water was also well represented in most regional models (AAIW; θ= [7-8] ◦C; S = [<35.4] and at γ= [27.38-27.92] kg/m3).420 In terms of the different NEA sub-regions, the best correlation was found for the AF sub-region (Fig. 10a), with r2∼0.99, for temperature and r2 ∼0.97, for salinity. As expected the lowest correlation is found in MWO sub-region (Fig. 10b) where the r2are 0.94 and 0.86 for temperature and salinity, respectively. Despite the fact that CA and MA sub-regions are better425 correlated with ARGO profiles, than the IP sub-region, it is important to note that the CA and MA sub-regions had less profiles than the IP region and thus measured less intrinsic variability. An overall statistical analysis was also performed for individual float trajectories in the whole NEA region. An example representative of these com-430 parisons is shown in figures 11 a 16 , where the largest Bias between ROMS (R C) solution and ARGO was often found in the sub-regions dominated by the MIW (between 1000 and 1700 m of depth) such as, AF (Fig. 11), MWO (Fig. 12) and IP (Fig. 13). However, better comparisons were found in regions away from the Gibraltar Strait namely: MA (Fig. 14), CA (Fig.435 15) and AA (Fig. 16). Generally, in the three experiments simulations the values of Bias are in the range of -2 and 2 ◦C for temperature and -1 and 1 for salinity for each ARGO float comparison. These results, suggest that the poorest representation of the high salinity Intermediate Mediterranean Water in the model, i.e. the salinity was underestimated in ROMS.440 20 The reason for these large differences are not in the number of sigma levels, since simulations with different numbers of sigma-levels were also considered, with no significantly different results. Nevertheless, there is a need to continue these experiments with coupled atmospheric-ocean models, considering a more realistic interchange between the atmosphere and the ocean. Haid-445 vogel et al. (2000) suggests that without a continuous source of water with S>36.5, T>11.8 ◦C and ◦=27.9 kg/m3, a significant freshening takes place during the model configuration, so the apparently better representation of MIW. Nevertheless, these corrections are often applied in longer model runs (+10 years), and thus apparently not so relevant for our 4 year NEA study.450 Peliz et al. (2007) and Mason et al. (2011), also prescribe the MIW outflow in the Gulf of Cadiz region, although this might be an acceptable band-aid parameterization, when using OGCM boundary conditions this might in fact compromise the initial and boundary conditions and thus introducing unexpected variability. To account for the inter-decadal variability of the MIW455 outflow, perhaps a better solution would be to reproduce such variability at the OGCM level, which would then propagate this onto the regional model solutions. 3.2.4. Case study of an IPUS episode (Aug 2005) An Iberian upwelling event was detected with an 8-day MODIS-Aqua460 composite of surface derived chlorophyll-a (Fig. 17a), for the 20 of August 2005. The same episode was also captured in the AVHRR derived SST map, as shown in the figure 17b. For the same period, ROMS calculated SST (R NQ) (Fig. 17c, d and e) showed similar low-temperatures suggesting the development of an upwelling event, accompanied by the formation of an465 21 upwelling filament, in the Extremadura promontory (between 38.5 and 39.5 ◦N, 17c). The same event was previously studied by Meunier et al. (2010) using SST and chlorophyll maps. Peliz et al. (2003) observed that in such events eddy shedding might occur as a result of the interaction between the flow with the topography discontinuities, like the Extremadura promontory,470 coherent with the location of some SST fronts. Recently, Batteen et al. (2007) identified three main reasons to explain the generation of filaments which include: (i) the baroclinic instability of the upwelling front; (ii) the effect of capes and promontories; (iii) planetary beta effect and bottom topography. In fact, all our ROMS simulations can reproduce lowest temperatures475 near the coast, suggesting the development of an IPUS event, however, R M and R NQ (Fig. 17c and d), computed the best results. The black box in the top panel of figure 8 highlights the strong upwelling event (see clorophyll-a map in figure 17a) for the VS buoy. The second black box represents the data collected by the CS buoy, also denoting another IPUS event (August, 2006).480 These results show that the experiments including adequate atmospheric forcing, are capable of reproducing the inter-annual variability measured by the buoys. The same can not be said for the climatological experiment R C which maintains a constant temperature, during both IPUS events. Quantitatively, the lowest spatial RMSE was found using R M, with an average485 value of 0.73 ◦C for the whole month of August, instead of 1.06 ◦C and 1.27 ◦C for R NQ and R C, respectively. As is expected, Bias values are also lower for the R M (0.42 ◦C) and for the R NQ (0.65 ◦C), compared to the R C experiment (0.91 ◦C). 22 4. Conclusions and suggestions for future works490 1 - The results from this work show the importance of atmospheric fluxes to generate a realistic regional ocean model solution for the NEA. There are also some advantages on downscaling an OGCM solution into a regional model to study the dynamics of the mesoscale processes and not relying only on climatological forcing. Current OGCM solutions of eddy-resolving495 (1/12◦), are only available from 2009 to the present. The kinetic energy spectra (Fig. 4) shows that ROMS seems to reconstitute the theoretical spectrum of oceanic mesoscale, representing the slopes k−5/3and k−3, according to Kolmogorov turbulent cascade energy. The effective resolution of OGCM is at scales of ∼500km for SODA, ∼600km for ECCO and ∼100km500 for Mercator, where the energy levels drops down the slope of k−3. 2 - The statistical analysis show a good data-model comparisons, with SST and EKE derived from satellite data, except in coastal areas where the satellite detection capability is limited. The better correlations are found in offshore regions. Is equally important to note that model-data comparisons505 with AVHRR derived SST, showed better statistical agreement, than when comparing model with MW SST data. Model comparisons with in-situ buoy data, R M and R NQ reproduced well the daily sub-surface variability (Fig. 8). 3 - ROMS numerical experiments also replicate, rather well, the detec-510 tion the eddy corridors documented in Sangr`a et al. (2009), using altimetry data extracted from AVISO. This eddies propagate westward as originally proposed by Pingree (2002) and recently confirmed by Sangr`a et al. (2009). 4 - The water column composition is best represented using ROMS forced 23 with climatological data. The central and deep waters, represented by R C515 are in good agreement with the ARGO profiles, whereas the depth of occurrence of the Mediterranean Intermediate Water (MIW) is not as accurately represented, since ROMS underestimates the highest salinity values. 5 - The analysis of the results also show that ROMS forced with high temporal resolution atmospheric fluxes (experiments R NQ and R M), re-520 produces well an IPUS event, (20 of August of 2005). Realistic upwelling filaments are also well simulated, representing the interaction of the flow and the local topography, along the Iberian Peninsula coast. Future works include, but are not limited to, the use of coupled a oceanatmospheric mode, in order to study the surface dynamics without compro-525 mising the water masses representativity. It is also expected that the inclusion of tides and river outflows, will contribute for a better representation of the salinity values, in particular the adequate location of MIW. 5. Acknowledgments The authors wish to acknowledge funds from projects: RAIA (0313 RAIA 1E)530 and PROMECA (CTM2009-06993-E/MAR). Numerical model solutions were calculated at CIIMAR HPC unit, constructed using funds the FCT-Portuguese National Science Foundation pluriannual, and further improved using funds from RAIA.co project, co-funded by INTERREG-IV and by FEDER (‘Fundo Europeu de Desenvolvimento Regional, 2007-2013’), through the POCTEP535 regional initiative. The altimeter products were produced by SSALTO/DUACS and distributed by AVISO with support from CNES. MODIS data was extracted using the online system, developed and maintained by the NASA. 24 Pingree, R., Sinha, B., 2001. Westward moving waves or eddies (storms) on the subtropical/azores front near 32.5n? interpretation of the eulerian currents and temperature records at moorings 155 (35.5w) and 156 (34.4w). Journal of Marine Systems 29 (1-4), 239–276.675 Pollard, R., Griffiths, M. J., Cunningham, S. A., Read, J. F., Prez, F. F., Ros, A. F., 1996. Vivaldi 1991 - a study of the formation, circulation and ventilation of eastern north atlantic central water. Progress in Oceanography 37 (2), 167–192. Relvas, P., Barton, E. D., Dubert, J., Oliveira, P. B., Peliz, A., da Silva,680 J. C. B., Santos, A. M. P., 2007. Physical oceanography of the western iberia ecosystem: Latest views and challenges. Progress in Oceanography 74 (2-3), 149–173. Reynolds, R. W., Smith, T. M., 1994. Improved global sea surface temperature analyses using optimum interpolation. Journal of Climate 7 (6),685 929–948. Reynolds, R. W., Smith, T. M., Liu, C., Chelton, D. B., Casey, K. S., Schlax, M. G., 2007. Daily high-resolution-blended analyses for sea surface temperature. Journal of Climate 20 (22), 5473–5496. Sangr`a, P., Pascual, A., Rodr´ıguez-Santana, A., Mach´ın, F., Mason, E.,690 McWilliams, J. C., Pelegr´ı, J. L., Dong, C., Rubio, A., Ar´ıstegui, J., Marrero-D´ıaz, A., Hern´andez-Guerra, A., Mart´ınez-Marrero, A., Auladell, M., 2009. The canary eddy corridor: A major pathway for long-lived eddies 31 in the subtropical north atlantic. Deep-Sea Research Part I: Oceanographic Research Papers 56 (12), 2100–2114.695 Shchepetkin, A. F., Mcwilliams, J. C., 1998. Quasi-monotone advection schemes based on explicit locally adaptive dissipation. Monthly Weather Review 126 (6), 1541–1580. Shchepetkin, A. F., McWilliams, J. C., 2003. A method for computing horizontal pressure-gradient force in an oceanic model with a nonaligned verti-700 cal coordinate. Journal of Geophysical Research C: Oceans 108 (3), 35–1. Shchepetkin, A. F., McWilliams, J. C., 2005. The regional oceanic modeling system (roms): A split-explicit, free-surface, topography-followingcoordinate oceanic model. Ocean Modelling 9 (4), 347–404. Song, Y., Haidvogel, D., 1994. Numerical simulations of the ccs under the705 joint effects of coastal geometry and surface forcing. pp. 216–234. Stauffer, D. R., Seaman, N. L., 1990. Use of four-dimensional data assimilation in a limited-area mesoscale model. part i: experiments with synopticscale data. Monthly Weather Review 118 (6), 1250–1277. Vallis, G. K., 2006. Atmospheric and Oceanic Fluid Dynamics. Cambridge710 University Press. Volkov, D. L., Fu, L. L., 2010. On the reasons for the formation and varibiality of the azores current. Journal Geophysical Oceanography 40, 2197–2220. 32 Figure 1: Eddy Kinetic Energy (EKE) computed from AVISO data (average 2003-2006 years) for NEA domain. The black boxes denote the critical sub-regions used for the solution validation, where: AF-Azores Front; MWO-Mediterranean Water Outflow; IPIberian Peninsula; MAMadeira Archipelago; CA-Canary Archipelago; and AA-Azores Archipelago. Dashed red lines show the sections used to plot of the H¨ovmoeller diagrams. 33 500 450 400 350 300 250 200 150 100 50 Figure 2: Temporal series of Sea Surface Temperature (◦C) for each OGCM (black line) compared with SST from AVHRR (blue line) and MW SST (red line) for the period of time since 2004 to 2006 where (a) is ECCO model; (b) SODA model; (c) HYCOM model and (d) Mercator model. One can note the differences in their temporal resolutions. 34 Figure 3: (a) Depth average MAE of temperature profiles between ARGO and ROMS profiles (0 to 2000 m averaged) for the 2003 to 2006 period of study. (b) MAE for salinity profiles. 35 Figure 4: Kinetic energy spectra (m3/s−2) of three OGCM models: Mercator (1/4 degree; gray dashed line); SODA (1/2 degree; gray dot line); ECCO (1 degree; gray dasheddotted line); and the three ROMS simulation experiments: R M run (1/12 degree; blue line); R NQ (1/12 degree; violet line); and R C run (1/12 degree; red line) for all domain at 35◦N Latitude and for one month of simulation in Jan-2003. The spectra show how ROMS seems to reconstitute the theoretical spectrum of oceanic mesoscale represented by the slopes -5/3 (solid black line) and -3 (dashed black line). 36 Speclral densily 10' ~-'-'~--------~~'-------'----'---'---'--'-'-'-'----------- -'-------'----'--r - == - ~ - M~E~R~ 0.001 0.01 Wavenumber km1 """' SODA _.- ECCO - R_M - R_C -R_NQ Figure 5: Temporal series of Root Mean Squared Error (top panel), Bias (middle panel) and correlation coefficients (bottom panel) for the whole area of NEA and during the four simulated years (2003-2006), between the three experiments made for this study and SST from AVHRR, where blue (o) are R M, violet (x) are R NQ and red (+) are R C. 37 o R_M x R_NQ + R_e 1.9 w (/) :; o: Mac Juo Sep 2003 M ac Juo Sep 2004 Mae Jun Sep 2005 Mae Jun Sep 2006 Mae Juo Sep 2003 Mae 0.95 ';,: 0.9 0.85 0.8 Mar Jun Sep 2003 Ma, Jun Sep 2004 Mar Jun Sep 2005 Mar Jun 2006 Time (daily) 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN r2 SST (ROMS−AVHRR) 0.75 0.8 0.85 0.9 0.95 1 Figure 6: (a). Temporal averages of correlation coefficient (r2), for four years since 2003, comparing SST from R M experiment results with SST from AVHRR satellite data, computing for each sub-region. 38 , " . . 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN 30oW 24oW 18oW 12oW 6oW 28oN 32oN 36oN 40oN 44oN RMSE EKE (ROMS−AVISO) 100 200 300 400 500 600 700 Figure 7: Temporal averages of Root Mean Squared Error (RMSE) for four years since 2003, between EKE from R M experiment and AVISO altimeter data, computed to each sub-region. Units are in cm2/s2 39 Figure 8: Temporal series during 2003-2006 of temperature at 3 m depth from Puertos del Estado buoys (black dashed lines): Villano Sisargas (VS), Cabo Silleiro (CS) and Estaca de Bares (EB) compared at the same depth with R M (blue line), R NQ (violet line) and R C (red line). Mean values of correlation coefficients are also showed in the top of these figures, higher values were archived for R M and R NQ than for the R C run. Black boxes correspond to an IPUS event of August 2005 for VS buoy and other at August 2006 for CS buoy. Note that the temporal axis are not always the same due the miss data in some days of the buoy. The location of the three Puertos del Estado buoys are also shown in the bottom panel with the bathymetry of the area. 40 Figure 15: ARGO buoy number 6900506 in the CA sub-region. 47 -500 -1000 -1500 -2000 09/24/06 -500 -1000 -1500 -2000 09/24/06 Vertical section lor bouy 6900506 -Bias 01 Temperature (10 proliles) 10/04/06 10/04/06 10/14/06 10/14/06 26°N 10/24/06 11/03/06 11/13/06 Bias 01 Salinity 10/24/06 11/03/06 11/13/06 1-- ARGO -- ROMS 1 " , 04 ." .• 25°N '-==--~=~- 18°W 16°W 11/23/06 11/23/06 12/03/06 12/13/06 0.5 -0.5 -1 12/03/06 12/13/06 Figure 16: ARGO buoy number 6900166 in the AA sub-region. 48 Vertical section lar bouy 6900166 -Bias 01 Temperature (61 proliles) -200 r - .., r""'"" --- ~- ""!!!I • • ~ -400 -600 -800 -1600 -1800 , ! 12/07/03 12/17/03 12/27/03 01/06/04 01/16/04 01/26/04 Bias 01 Salinity -200 -400 0.5 -600 -800 -1000 -1200 -1400 -05 -1600 -1800 -1 12/07/03 12/17/03 12/27/03 01/06104 01/16/04 01/26/04 1-- ARGO -- ROMS 1 400N r¡==~~=====~~ ... • ..~ . Figure 17: (a) Map of chlorophyll-a concentration (log10*100 mg/m3) from MODIS for 8 days averages (13 to 20 August 2005) indicating an IPUS event in the Iberian Peninsula coast. (b) SST ◦C from AVHRR at 20 Aug 2005 show the same upwelling event. (c), (d) and (e) maps of SST ◦C from the experiment R NQ, R M and R C, respectively, for the same IPUS event of 20 Aug 2005. 49 Chl-a (MODIS) SST (AVHRR) 27 440N 44 'N b 26 25 42 0N 42'N 24 2.5 23 40 0N 40 'N 22 21 3S0N 38 'N 20 1.5 19 36 0N 36 'N 18 1S0W 1S0W 12"w gOW SOW 17 SST (R_NO) SST(R_M) 27 27 44 0N 2. 44 0N 2. 25 25 42 0N 2. 420N 23 23 400N 22 400N 21 2 1 3S0N 20 380N 20 l. 1. 36 0N l. 360N " 17 17 SST(R_C) 27 44 0N 26 25 42 0N 2' 23 400N 22 21 3S0N 20 l. 36 0N l. 17 Table 1: Sources of data used in this study for the validations and comparison with the different ocean models 50 Table 2: Mainly characteristic of different OGCMs used in this study 51 Table 3: Temporal means of statistic parameters: MAE (Mean Absolute Error); RMSE (Root Mean Squared Error) and Bias of EKE for each OGCM model vs EKE from AVISO data. Grey shaded show the OGCM with better characteristic according to values of metrics validations. 52 Table 4: Validations made over temporals means for the whole area of NEA between the three ROMS experiments and SST satellite data from AVHRR and MW SST. 53