**1. Introduction**

Operational oceanography, as defined by EuroGOOS (European Global Ocean Observing System [1]) is the activity of systematic and long-term routine measurements of the ocean and atmosphere, and their fast interpretation and dissemination. Operational Oceanography is rapidly maturing, and its capabilities are being enhanced, mostly in response to a growing demand of regularly-updated ocean information. End-users and stakeholders are key to sustain operational oceanographic systems, as well as to foster new ones.

A complete list of the range of services and a description of how end-users and stakeholders use these operational services goes beyond the scope of this introduction (see Schiller et al., 2018 and Davidson et al., 2019 for more extensive reviews [2,3]). However, in general, primary sectors supported by operational oceanographic services are those related to improve the safety and efficiency of marine activities. Many of the existing operational oceanographic services focuses on regional coastal waters, where most of

**Citation:** García-León, M.; Sotillo, M.G.; Mestres, M.; Espino, M.; Fanjul, E.Á. Improving Operational Ocean Models for the Spanish Port Authorities: Assessment of the SAMOA Coastal Forecasting Service Upgrades. *J. Mar. Sci. Eng.* **2022**, *10*, 149. https://doi.org/10.3390/ jmse10020149

Academic Editor: Christos Tsabaris

Received: 3 December 2021 Accepted: 19 January 2022 Published: 24 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

the protection of marine environments in sensitive habitats and human activities occurs (De Mey-Frémaux et al., 2019 [4]).

Among coastal activities, the ones related to port activity are certainly important due to the role that ports play in global and national economies. Note that the total gross weight of goods handled in EU (European Union) ports was estimated above 3.8 billion tonnes in 2015. The Spanish Port System contributes to this activity, handling 447 million tonnes during 2015, with a 4.5% of annual growth in the 2010–2015 period (EU EuroStat 2017 [5]). Approximately, 85% of total Spanish imports and 60% of exports are channelled through ports.

However, the ports are highly impacted by extreme met-ocean drivers (especially wind, waves, surface currents and sea level). Physical environment constrains Port equipment and their activities during all phases of Port life (from its design phase, in the planning tasks and during daily operations). Thus, oceanographic historical or reanalysis data are needed to be used in design criteria for offshore structures. Coastal infrastructure and maritime industries also require short-term high-resolution forecasts of wind, wave and current to manage their operational decisions. At a wider temporal horizon, longer-term forecasts and climate projections are needed for strategic decisions, port adaptation and planning. In order to answer to these complex challenges, the Spanish Port System launched SAMOA (Sistema de Apoyo Meteorológico y Oceanográfico a las Autoridades portuarias—System of Meteorological and Oceanographic Support for Port Authorities; Álvarez Fanjul et al., 2018 [6]). From 2014, Puertos del Estado (PdE) has led this SAMOA initiative developing and providing high-resolution met-ocean operational products to feed Decision Support Systems of the Spanish Port Authorities.

This SAMOA initiative combines met-ocean observation, ocean modelling and enduser service tools. One of the SAMOA components (the one referred in this work) is focused on the design and implementation of very high-resolution ocean circulation forecast systems, able to solve from coastal to very local scales within the ports. Sotillo et al., (2019) [7] provides an extensive description of this SAMOA integrated coastal/port forecast service. This system solves governing equations for ocean currents, sea level, temperature, salinity and concentrations of tracers; as one of the building blocks of any operational oceanographic systems aimed to forecast physical coastal processes.

In that sense, in the last decade, a grea<sup>t</sup> variety of operational ocean models were used, mostly at global (Lellouche et al., (2013, 2018), Liu et al., 2021 [8–10]) and basin scales on European Seas, based on different model codes (with different physical parameterization sets), spanning a wide range of spatial and temporal scales, using different forcing data sets and relying on data assimilation methods (Bell et al., 2009; Bell et al., 2015; Tonani et al., 2015 [11–13]). Today, the Copernicus Marine Service (CMEMS; https:// marine.copernicus.eu/ accessed on 18 January 2022 [14]), focused on global and regional scales in European basins, is well-established and the sustained availability of its "core" operational ocean products has favoured the proliferation of "downstream" services devoted to coastal monitoring and forecasting (Letraon et al.; 2019 [15]). Capet et al., (2019) [16] in their review of current European capacity in terms of operational marine and coastal modelling systems, map 49 organizations around Europe delivering 104 operational model systems simulating mostly hydrodynamics, biogeochemistry and sea waves at regional and coastal scales. The PdE SAMOA coastal forecast model applications for Spanish ports, are nested into the CMEMS forecast solution for the Iberian-Biscay-Ireland (IBI) regional seas (Sotillo et al., 2015, Aznar et al., 2016 [17,18]), thus contributing to this European coastal model capacity.

In their detailed review on the basis of coastal modelling, Kourafalou et al., (2015) [19] pointed out that advancement of coastal ocean forecasting systems requires continuous scientific progress in several key topical areas. Among them, the following research lines were identified as priority ones to evolve the SAMOA service: (i) to understand primary mechanisms driving coastal circulation; (ii) to develop adequate methods to dynamically embed coastal systems in larger scale systems; and (iii) to include in downscaled solutions methods to adequately represent air–sea and land interactions, involving atmosphere-waveocean coupling. Dealing with these primary science topics was fundamental to face the initial SAMOA development phase (2014–2017) that resulted in nine new SAMOA coastal model applications (running operationally since 2017), as well as to align the research needed to improve the SAMOA circulation models. This second development phase, the SAMOA-2 Project (2019–2021), allowed to extend the SAMOA forecast service to a higher number of Spanish Ports (31) and to upgrade the ocean model applications in which the SAMOA service is based.

This contribution presents the research conducted to develop the next generation of SAMOA coastal circulation model applications. This research has focused on improving the SAMOA forecasts through three main objectives:


A complete description of the sensitivity tests performed to improve the coastal model that support the SAMOA operational service, joint with the quality assessments via extensive model validation (using different local available in-situ and remote sense observations), is herein provided.

The paper is organised as follows: Section 2 outlines the PdE SAMOA initiative, describing its objectives, components and major benefits as well as the amelioration axes followed in its currently on-going second development phase (SAMOA-2). Section 3 describes the model sensitivity tests performed aimed to improve the SAMOA model solution. Section 4 presents a detailed discussion of the results with an analysis of the proposed SAMOA updates in the different port domains. Finally, Section 5 summarises main conclusions and gives a look ahead to on-going and future research topics that may enhance SAMOA operational forecast capabilities.

#### **2. The Operational Context: The PdE SAMOA Coastal Forecast Service and Its Second Development Phase (SAMOA-2)**

Since the mid-1990s, PdE has acted as an operational oceanographic centre, sustaining an integrated multiplatform ocean monitoring and forecasting system that provides service from regional to coastal scales. Met-ocean parameters, such as wind, waves, sea-level and currents have been extensively monitored and forecasted. Main applications range (i) the safety of port operations, (ii) impacts on ship draught allowance, (iii) water quality and (iv) navigation and piloting activities.

The SAMOA initiative, co-financed by PdE and the Spanish Port Authorities, was born in 2014 as an answer to the complex port needs in terms of coastal and local met-ocean information (Álvarez Fanjul et al., 2018 [6]). SAMOA aims to enhance the PdE traditional operational oceanographic service, implementing new operational systems for each Port Authority, with tailor-made operational met-ocean information, as well as some added value products, previously requested by each Port Authority.

The SAMOA service is organized as a modular infrastructure, with some of the SAMOA modules focused on enhancing near-real-time monitoring (i.e., by means of local upgrading of current PdE observational networks with specific instrumentation), whereas some others are focused on implementing new local high-resolution forecast applications for atmosphere, waves and ocean circulation. The enhancement of PdE model capabilities within SAMOA, and its follow-up in SAMOA-2, has been certainly significant, being a total of 17 new high-resolution atmospheric models (1 km resolution, based on HARMONIE model), 20 wave agitation models (5 m resolution mild slope model applications) and 31 circulation models (70 m ROMS model set-ups) developed and operationally implemented. The SAMOA modular architecture allows each Port Authority to decide what specific SAMOA services are needed to be implemented in their ports, mainly addressing their specific needs for met-ocean information. A total

47 ports from 28 Spanish Port Authorities benefited from these new SAMOA modelling and monitoring capabilities.

To properly exploit new SAMOA added-value products, a specific tool was developed for the Ports: The CMA (Cuadro de Mando Ambiental or Environmental Control Panel). This CMA dashboard is a tailor-made web service that provides easy and customized access to operational met-ocean information in the ports, as well as advanced viewing capabilities (see snapshot in Figure 1a) and other end-user services (such as an alert system (through email/SMS), or a tool to automatically generate daily bulletin reports with environmental conditions at selected locations). Managers in the ports gran<sup>t</sup> access to this CMA tool and they can define different levels of user permissions. A growing community of 1913 port registered users benefits from the CMA tool.

With respect to ocean circulation forecast, and after running regional ocean circulation model applications for more than a decade (i.e., the Spanish ESEOO system; Sotillo et al., 2008 [20]), PdE does not run anymore a regional forecast component, relying since 2014 for these scales on the Copernicus Marine IBI-MFC regional forecast service (Aznar et al., 2016 [18]). This use of Copernicus regional products (currently delivering forecast products at 1/36◦ resolution that covers all the Spanish waters) has allowed PdE to focus their ocean modelling resources on running very high-resolution models at specific hot spots (p.eg. the Gibraltar Strait and nearby areas, operationally covered by the PdE SAMPA forecast service; Sanchez Garrido et al., 2013 and Sotillo et al., 2016 [21,22]) and on developing the new SAMOA operational downscaled systems for ports, here described.

The related PdE ocean circulation forecast products, updated on a daily basis, can be downloaded through the PdE catalogue (https://opendap.puertos.es/thredds/catalog. html accessed on 18 January 2022 [23]) and visualized through the mentioned CMA Dashboard and the PdE PORTUS web page https://portus.puertos.es/ accessed on 18 January 2022 [24] (see Figure 1b shows a snapshot of the modelled surface current field forecasted for coastal waters nearby the Bilbao harbour).

In SAMOA-2, it was decided to enhance the coastal/port circulation module by adding an extra oil spill forecast service linked to the implementation of a local high-resolution circulation model. This new option allows to have an on-demand use of the PdE oil spill forecast service (based on MedSlik model, De Dominicis et al., 2013 [25]) and to visualize their model outputs through the CMA dashboard (see snapshot of the output tool in Figure 1c). This added oil spill module played a relevant role on fostering the interest of Port Authorities in high-resolution coastal circulation products for their coastal fringes. This Ports' interest on currents, and mainly on its advective and dispersive effects, is partially explained by the legal commitment that Port Authorities need for ensuring good water quality within harbours (European Parliament, Directive 2019/883 [26]).

In the first SAMOA development phase, nine high-resolution coastal circulation systems were implemented for different Spanish ports located in the Mediterranean, the Iberian Atlantic and the Canary Islands (see green dots in Figure 2). This first SAMOA implementation of high-resolution circulation systems was performed thanks to the collaboration between PdE and the Maritime Engineering Laboratory of the Polytechnic University of Catalonia (LIM/UPC). The same team has continued along SAMOA-2 with the R&D works needed to face the challenge of (i) improving currently existing SAMOA port circulation systems and (ii) to extend this SAMOA forecast service to other requested ports (i.e., multiplying by three the number of Spanish ports where SAMOA coastal forecasts are implemented by 2021).

**Figure 1.** 3 Example of Port user-oriented viewing capability through the SAMOA CMA environmental dashboard: (**a**) 3-days forecast information for wind, waves and currents at Algeciras Bay; Depicted maximum forecasted values and coloured end-user alert symbols (green-yellow-red) following specific user-customized alert thresholds at location of interest for the Port operations; (**b**): Map of SAMOA high-resolution forecasted surface current field in nearby waters of the Bilbao Port. (**c**) Example of Oil spill forecast in front of Barcelona Port. Outputs from the PdE Oil Spill forecast service using SAMOA high resolution forecasted currents.

**Figure 2.** SAMOA Coastal/Port circulation forecast systems in operations. Green dots: Operational models systems developed within SAMOA1 (2014–2017). Red dots: Operational model systems developed within SAMOA2 (2018–2021). Bathymetry (in meters) is depicted. Note that systems in the Canary Islands are shown in the bottom left corner box (SW limit: 18.5◦ W, 27.5◦ N; NE limit: 13.2◦ W, 29.5◦ N).

This noticeable increase in the number of SAMOA Port systems (red dots in Figure 2 shows location of the Spanish ports where new circulation systems are deployed after the SAMOA second development phase) results in an almost complete coverage of the Spanish coast with high-resolution coastal circulation forecasts.

The operational SAMOA local circulation forecast systems daily produce short term (+3 Days) forecasts of 3-D currents and other oceanographic variables, such as temperature, salinity, together with sea level. Each SAMOA system uses the ROMS model (Shchepetkin and McWilliams (2005); present code source and documentation at the ROMS website [27,28]), and consists of two nested regular grids with spatial resolution of ~350 m and ~70 m for the coastal and harbour domains, respectively. The model is forced with very high-resolution atmospheric forcing (provided by the Spanish Meteorological Office AEMET) and it is nested into the core Copernicus IBI regional solution (IBI\_PHY hereafter). Figure 3 shows a schematic view of the SAMOA system; for an end-to-end description of the SAMOA model set-up, see Sotillo et al., 2019 [7].

**Figure 3.** Architecture of the SAMOA circulation forecast system. (**a**) Operational scheme, detailing (i) input data sources, including atmospheric forcing and the ocean solution imposed at the Open Boundary Conditions (OBCs), (ii) the numerical core (ROMS) with domain organization, and (iii) delivered products. (**b**) SAMOA Operational Releases: The timeline shows main releases of this service since 2018 and their most prominent changes. Please note that the acronyms SAM\_INI, SAM\_ADV and SAM\_H3D will be used throughout the text, to name the different model set-up experiments. Likewise with IBI\_PHY (the SAMOA parent solution).

Product quality assessment is a key issue for any operational forecast service (Ryan et al., 2015 [29]). Nevertheless, one of the main bottlenecks identified in the recent review performed by EuroGOOS for European coastal operational ocean model services (Capet et al., 2019 [16]) is the lack of an adequate delivery in near-real-time (NRT) of operational observations. This lack of operational observations, especially on coastal areas, restricts the systematic exploitation of (i) the data assimilation capacities in operational marine modelling systems and (ii) the provision of NRT assessments of operational ocean model products. Note that only 20% of operational models currently available for European seas provide a dynamic uncertainty together with their forecast products.

In the case of SAMOA, an exhaustive NRT validation of the forecast products operationally generated is performed. This operational validation is based on a routine monitoring of the quality of the SAMOA forecast products on a daily and monthly basis. This SAMOA model forecast validation is using all operational ocean observation available in the port and nearby coastal waters. To this aim, an extension of the NARVAL tool (originally developed for the CMEMS IBI MFC regional model solutions (Lorente et al., 2019 [30])) was implemented to validate the SAMOA circulation model products. The comprehensive multi-parametric ocean model skill assessment is performed by using all available operational observational sources in the coastal domains. The list of observational data sources includes satellite L3 and L4 SST products, together with in-situ observations from moorings and tide-gauges, as well as HF-Radars. Further details about the complete list of observational products and platforms used by NARVAL to validate SAMOA can be seen in Sotillo et al., 2019 [7].

Another SAMOA module, consisting of observational campaigns at harbour waters was offered to Port Authorities. These campaigns include at the same time (and for a three month period) deployment of 1 ADCP current-meter, fixed (and repeated) CTD temperature and salinity stations, as well as additional meteorological stations. It is expected that this SAMOA extensive monitoring campaign will help the Ports to increase the knowledge on their waters and to validate the SAMOA met-ocean forecast applications. In that sense, it is remarkable that all Port Authorities (9) who have requested this SAMOA specific in-situ observational port campaign count with a SAMOA Port Circulation Forecast service. However, to assess the model sensitivity tests shown in the present work, only operational observational data sources were used (from PdE HF Radar, tide gauges and coastal or deep-water mooring stations), since data from the new SAMOA observational campaigns were not available when SAMOA model tests were performed.

#### **3. Methodology: Model Sensitivity Tests for Improving the SAMOA Circulation Forecast Services**

After more than 4-years running operationally the first 9 SAMOA port forecast services, joint with the continuous quality assessment of their model products (performed throughout the NARVAL validation tool described in the previous section), the following main concerns about the SAMOA solutions were identified:


Specific research was conducted with the aim to minimize these limitations via improving the SAMOA model set-up for the forthcoming releases.

External forcing, and particularly the atmospheric one, remains as one of the major issues restricting the accuracy of operational coastal systems. In this sense, improving external atmospheric forcing has been identified as a major concern for enhancing the reliability of high-resolution forecast services (Capet et al., 2019 [16]). Furthermore, the improvement of atmospheric forcing (and a methodology for a better usage in coastal domains) is highlighted in the SAMOA roadmap (outlined in Sotillo et al., 2019 [7]), as one of its three main research lines.

On the other hand, Kourafalou et al., (2015) [19] point out that coastal ocean forecasting systems requires continuous scientific progress in a primary science topic as it is the downscaling from larger scale models, with a need of ad-hoc nesting procedures. This includes assessment of the boundary conditions from the larger scale parent systems to the coastal nested systems, and refinement of the model set-up (including grids, topographic details, forcing data and schemes used as open boundary condition). A challenge for these appropriate nesting procedures is to ensure consistency in the fluxes between the downscaled coastal model and its coarser parent solution. This consistency will avoid problems such as triggering unrealistic gravity transient currents that may unleash spurious dynamical coastal model features.

Consequently, the following three model sensitivity tests were performed:


Table 1 summarises the main differences among the three different model sensitivity datasets. As denoted in Figure 3b, SAM\_INI corresponds to the former operational version of SAMOA (active since October 2018 to October 2019), whereas SAM\_ADV is the opera-

tional set-up currently in production. SAM\_H3D will be the forthcoming SAMOA set-up version (to be operationally released in 2022).

**Table 1.** Summary of the main differences among the sensitivity tests conducted. The columns denote the three SAMOA model configurations (SAM\_INI, SAM\_ADV and SAM\_H3D) and the rows the different upgrades. ✔ means included in each model set-up, and ✖ not included.


All the SAMOA model set-up novelties have been extensively tested. The proposed sensitivity model runs were validated using in-situ and remote sensing observations (i.e., from PdE coastal and deep-water mooring buoys for surface temperature, salinity and currents, tide gauges for sea level and HF-Radars for surface currents, available the latest only in some of the coastal model domains).

A complete description of the different sensitivity model tests performed to improve the operational coastal model applications that sustain the SAMOA forecast service for Spanish ports is provided below. Results from the different model tests are provided in the next section.

#### *3.1. Upgrade of the Atmospheric Forcing: The Initial SAMOA-2 Model Configuration (SAM\_INI)*

SAM\_INI is the first model configuration that is used in the sensitivity test. SAM\_INI set-up is closely related with the SAMOA-1 set-up (Sotillo et al., 2019 [7]), except for a key issue: SAM\_INI uses as atmospheric forcing the new AEMET HARMONIE data (Bengtsson et al., 2017 [31]), rather than the deprecated HIRLAM forcing data (also provided by the Spanish Met Office AEMET) that were used in the SAMOA-1 service. As shown in the operational scheme depicted in Figure 3, HARMONIE is used for the first 48 h of forecast, then completing the forecast range between 48 to 72 h with the ECMWF-IFS data (ECMWF 2019 [32]) provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). The HARMONIE data have a spatial resolution of 2.5 km and an hourly frequency, whilst the ECMWF-IFS forcing have a resolution of 10 km and tri-hourly frequency. It should be noted that ECMWF-IFS, with a global coverage, is the parent solution for the AEMET HARMONIE. The latter is a limited area-system with two domains: (i) a first one that spans the whole Iberian Peninsula and Balearic Islands, and (ii) a second one that covers the Canary Islands.

For both atmospheric products, the following atmospheric variables are used: wind fields, sea-level pressure, 2-m air temperature, 2-m relative humidity, precipitation, latent heat fluxes and sensible heat fluxes. In this first SAM\_INI model configuration, the atmospheric fluxes are computed with the same methodology presented in Sotillo et al., 2019 [7]).

As pointed out in Table 1, the SAM\_INI model configuration was used in the SAMOA operations since October 2018, being the SAMOA operational set-up for one year until the last release (held in October 2019).

#### *3.2. Upgrade of the Open Boundary Condition (OBC) Scheme and Bulk Fluxes: The Current Operational SAMOA Set-Up (SAM\_ADV)*

As mentioned above, the NARVAL validation system led to detect inconsistencies in the surface current fields forecasted by the SAMOA-1 systems and the parent solution (IBI-MFC). Additionally, the update of the atmospheric forcing in the SAM\_INI configuration also intensified error metrics in the Sea Surface Temperature (SST).

With the aim of minimizing some of these identified operational shortcomings, a new SAMOA model set-up (termed SAM\_ADV in Figure 3, and hereafter) was proposed. This new SAM-ADV configuration had two main objectives:


The SAM\_INI set-up for the OBC treatments was featured by:


However, the SAM\_ADV set-up proposed the following changes:


This new OBC treatment was proposed to ensure that, both normal and tangential, components of the velocity and passive tracers were conserved from the IBI\_PHY parent data to the SAMOA solution. Note that the OBC scheme used in the SAM\_INI set-up (and in the SAMOA-1 one) provided systematic inconsistencies with the tangential component of the velocities. These boundary inconsistencies affected the SAMOA solutions most especially in mesotidal and macrotidal regimes (that means in the SAMOA systems implemented in the Canary Islands and in the Cantabrian Sea, respectively).

The performance of this new SAM\_ADV configuration (and the difference with the SAM\_INI one) was tested in all systems by a one-month simulation period (June 2019). The time interval, though limited, was enough to highlight the benefits of this proposed upgrade.

#### *3.3. Use of a Higher Temporal Frequency Imposed Data at the Boundary and Upgrade of the Model Physics: The Forthcoming SAMOA Operational Set-Up (SAM\_H3D)*

Despite that the SAM\_ADV set-up remarkably enhanced the consistency between the SAMOA and IBI\_PHY at the boundaries; in some specific cases (i.e., for the Gran Canaria and Almería systems) the SAMOA performance was not satisfactory enough. A possible reason may come from the mismatch among the temporal frequency of the different variables in the OBCs. For instance, in previous set-ups (SAM\_INI and SAM\_ADV), sealevel and barotropic currents are updated on an hourly basis; but the total velocities, temperature and salinity have daily-mean frequency. Such treatment, usual in downstream services, may lead to non-conservation issues of the momentum and passive tracers.

The Copernicus service recently started delivering a new IBI forecast 3D hourly. This new product comprises hourly fields of total velocities, temperature and salinity along the whole water column for coastal and shelf areas of the IBI domain. The delivery of this new CMEMS IBI product means an opportunity for SAMOA to improve its downscaling. Hence, its use in the SAMOA nesting has been extensively tested in this SAM\_H3D experiment, aimed to provide insights on the forthcoming SAMOA release.

This new nesting has been tested in five SAMOA systems: three microtidal Mediterranean ones (Barcelona, Tarragona and Almería), and two Atlantic ones (a mesoscale one, Gran Canaria, and a macrotidal one: Bilbao). The SAM\_H3D testing period ranges one year (from October 2019 to October 2020), in order to assess how the SAMOA systems would behave throughout different season conditions. The OBC treatment in SAM\_H3D is analogous to the one used in the SAM\_ADV (previously described), but adapted to the higher temporal frequency of the IBI\_PHY imposed data.

Additionally, specific changes in the physics have been implemented in the operational ROMS version (3.7) that currently uses the SAMOA system, namely:


These changes aim to provide consistent surface fields (be it heat fluxes or surface stresses), by including last advances in air-sea interaction processes.

#### *3.4. Evaluation Criteria and Error Metrics*

From each of the three SAMOA model test experiments (in the different port domains selected in each case), it was produced a daily set of daily-averaged and hourly-averaged data. In this contribution, the model validation was focused on the hourly-averaged products, because they are more suitable to assess two main issues:


The hindcasted hourly-averaged products (the ones produced in the first 24 h of each forecast cycle; see Figure 3a) have been compared with hourly-averaged observations from the PdE Monitoring Network. This network comprises deep water and coastal buoys (REDEXT and REDCOS, respectively [40,41]), tidal stations (from the REDMAR network [42]) and HF Radar [43]. All these measurements have passed state-of-the-art Quality Control processes and can be accessed in NRT basis from the PORTUS [24] website.

Performance of the different SAMOA model tests have been assessed by comparing the different models with the available observational datasets at each coastal and port domain (see in Table 2 the observational coverage available at each SAMOA domain).

The following hourly-averaged error metrics has been computed for each dataset in each domain (when available observations): mean bias, correlation, Root-Mean Square Error (RMS) and Coefficient of Efficiency (COE). The metrics that will be shown in Tables 3–5 are representative of the whole analysed period in each sensitivity test. Metrics in Section 4.1 were computed from time-series enclosing June 2019; whereas those for the Section 4.2 range from October 2019 to October 2020.

The Coefficient of Efficiency (COE) (Legates and McCabe, 1999, 2013 [44,45]) is obtained as (Equation (1)):

$$COE = 1 - \frac{\sum\_{i=1}^{N} |O\_i - P\_i|}{\sum\_{i=1}^{N} |O\_i - \overline{O}|},\tag{1}$$

where *Pi* and *Oi* refer to the computed and observed signals respectively, *N* is the number of time records and (¯) is the mean operator. A *COE* value of 1 means a perfect model. Despite *COE* having no lower bound, a *COE* value of 0 implies that the model is not more able to predict the measured values than the measured mean would do. Consequently, for negative *COE* values, the computed model signal would perform worse than the measured mean in predicting variations in the observed signal. For the sake of interpretation, please note that the *COE* and the skill score of Murphy (1988) [46] are both the same, in case the observed mean is used as the reference model in the skill score.

**Table 2.** Summary of the observational coverage available at each SAMOA domain. Rows denote the SAMOA domain, and the columns, the different elements of the PdE monitoring network (deep water buoys (EXT), coastal water buoys (CST), tidal gauge stations (TGS) and Radar HF) used in the model validation. The ✖ symbol means unavailability of a certain type of sensor in the area. If there exists a monitoring device within the SAMOA domain, its main features are summarized in brackets including: location (longitude, latitude), mooring depth (in m), spatial resolution (in km) for the HF Radars and measured variables (Sea Surface Temperature, SST; Sea Surface Salinity, SSS; Surface current speed, SC\_S; Surface current direction, SC\_D; Sea Level, SLev).


Despite error metrics being computed with hourly-data, monthly average metrics were also calculated for Figures 6 and 7. These monthly metrics were computed specially in the case of surface currents versus HF Radar data (some figures in the Result section show monthly observational and model outputs, together with the bias field). The aim of these specific cases was to demonstrate that SAMOA can reproduce the spatial signature of surface-currents with a certain skill.

The model validation assessment has been held with the SAMOA coastal domains (350 m horizontal resolution, see Figure 3a) for surface currents, temperature and salinity, measured by met-ocean sensors on coastal and deep-water buoys (moored outside the ports) and by HF radars (covering nearby coastal areas but not inside the harbours). However, sea level has been addressed with the SAMOA local Port model solutions (70 m horizontal resolution), because coastal tidal stations are deployed within the harbours. Another remark for the sea level assessment, is that mean value has been removed from all the time series (be it models or observations), because each model has a different vertical system of reference. Consequently, estimated biases may be misleading and they are not comparable. **Table 3.** Error metrics for the SAMOA SAM\_ADV and SAM\_INI test runs, plus the IBI\_PHY parent solution (computed with hourly observations). The metrics comprise the June 2019 period. Variables: SST (Sea Surface Temperature in ◦C), SSS (Sea Surface Salinity, in PSU), SC\_S (Surface current speed, in m/s), SC\_D (Surface current direction, in ◦), SLev (sea level, in m). The domain column identifies the SAMOA system. The Type denotes the sensor type used: TGS (Tide Gauge Station), CST (Coastal Buoy), EXT (Deep-water buoy). N counts the size of the sample (conditioned by the availability of hourly observation). Each error metric (Bias, Correlation, Root-Mean-Square Error (RMS) and Coefficient of Efficiency (COE)) provided for each model solution (SAM\_INI (INI), SAM\_ADV (ADV) and IBI\_PHY(IBI)). The last row of each variable represents the average error metric of all the sensors given a specific test run. Bold numbers highlight the best performing dataset.


**Table 4.** Error metrics for the SAMOA SAM\_H3D and SAM\_ADV test runs, plus the IBI\_PHY parent solution (computed with hourly observations). The metrics comprise the time range from October 2019 to October 2020. Variables: SST (Sea Surface Temperature in ◦C), SSS (Sea Surface Salinity, in PSU), SC\_S (Surface current speed, in m/s), SC\_D (Surface current direction, in ◦), SLev (sea level, in m). The domain column identifies the SAMOA system. The Type denotes the sensor type used: TGS (Tide Gauge Station), CST (Coastal Buoy), EXT (Deep-water buoy). N counts the size of the sample (conditioned by the availability of hourly observation). Each error metric (Bias, Correlation, Root-Mean-Square Error (RMS) and Coefficient of Efficiency (COE)) provided for each model solution (SAM\_ADV (ADV), SAM\_H3D (H3D) and IBI\_PHY(IBI)). The last row of each variable represents the average error metric of all the sensors given a specific test run. Bold numbers highlight the best performing dataset.


**Table 5.** Surface current error metrics at Tarragona and Gran Canaria domains vs. HF Radar, at a subset of representative points. The metrics comprise the time range from October 2019 to October 2020. Variables: SC\_S (Surface current speed from HF Radar, in m/s), SC\_D (Surface current direction from HF Radar, in ◦). The domain column identifies each SAMOA system. The Pnt denotes the point ID shown in subpanel c in Figures 6 and 7. The N column counts the available observations (hourly resolution) that have been used for computing the metrics. Error metrics (Bias, Correlation and Root-Mean-Square Error (RMS)) provided for each model solution (SAM\_ADV (ADV), SAM\_H3D (H3D) and IBI\_PHY (IBI)). The last row of each variable represents the average error metric of all the sampling points given a specific SAMOA and test run. Bold numbers highlight the best performing dataset.

