1. Introduction
Modelling coastal systems particularly in regions of freshwater influence (ROFI), as it is the case of the Tagus estuary and its coastal adjacent area, is paramount to understand the physical and biogeochemical processes that characterize them. The simple measurement of environmental variables in situ alone or time-limited studies is not sufficient to translate and fully comprehended coastal processes since coastal systems are complex, decomposable, and large-scale [
1]. Indeed, the combination and interaction of the physical factors (bathymetry; coastline topography; estuarine outflow; heat flux through the sea surface) and the associated physical transport and dispersal processes of the coastal flow field, makes coastal systems complex and unique in its hydrodynamics [
2,
3]. Therefore, and like other large-scale systems, the coastal system must be partitioned or decomposed into a number of small systems and the best way to gain insights into coastal system structure, organization, and functioning is through the use of numerical models. These models, besides allowing a better understanding of the interactions between the various components of the system, are also indispensable to set scientific scenarios and to answer important scientific and management questions [
4]. Thus, only the proper knowledge of the processes that characterize the coastal systems, and their evolution in space and time, allow the design of robust coastal management plans. In this regard, hydrodynamic models assume a preponderant role since the establishment of currents and density fields is fundamental for transport models, whether they are Eulerian or Lagrangian, due to the importance of advection and/or diffusion in the study of issues related to the coastal system.
Among the numerical models that have been used to simulate the hydrodynamics of coastal systems (e.g., MARS3D [
5]; MIKE3 [
6]; SELFE [
7]; SHYFEM [
8]; TELEMAC [
9]; MIKE 21/3 [
10]; CSIRO [
11]; Delft3D [
12]), MOHID Water (Portuguese acronym of
MOdelação HIDrodinâmica) is one of the most widely used. The MOHID Water model was developed by MARETEC (Marine and Environmental Technology Research Center). MOHID Water is open-source and most relevant key strengths are its ability to: Deal with two-dimensional (2D) and 3D simulations, with sigma, cartesian or lagrangian vertical coordinates; deal with eulerian or lagrangian transport references; use the same biogeochemical formulations independently of the number of spatial dimensions or space reference; and incorporate alternative formulations for every process (due to its modular approach). The versatility and flexibility of the MOHID Water eases its implementation in any type of system and to accomplish different modelling requirements depending on the objectives of the work to be performed. Indeed, in the last four years the MOHID Water was applied in multiple studies worldwide, namely in Canada [
13], Colombia [
14,
15], Brazil [
16], Argentina [
17,
18], Uruguay [
19], Holland [
20], France [
21], Spain [
22], Croatia [
23], Australia [
24,
25], Malaysia [
26] and Korea [
27]. In Portugal, this model has been implemented in the entire Portuguese coast [
28,
29,
30] and in all main estuaries [
31,
32,
33,
34,
35].
Specifically, in the coastal system that comprises the Tagus estuary, most of the studies that used MOHID Water considered the coastal area and the Tagus estuary as sub-systems. In the case of the sub-system estuary, studies focused on sediment dynamics [
36,
37,
38], on chlorophyll and/or nutrients [
39,
40], on microbiology [
41,
42], on water level [
43], residence time [
44] and CO
2 fluxes [
45]. Concerning the sub-system coastal area, studies were directed to analyse the effects of the Tagus estuarine plume [
45,
46,
47].
All the above mentioned works regarded the sub-systems estuary and adjacent coastal area as different domains due to the spatial resolution that is required to represent local physical processes [
48,
49]. However, computational advances allowed to integrate both sub-systems into a single domain, which considers higher resolution grids in both horizontal and vertical directions, essential to increase the precision and accuracy of the coastal processes simulations [
50]. For this reason, only recently those two sub-systems were considered and modelled as an estuarine-coastal continuum [
47] using a variable grid in the horizontal axis, in order to have a higher horizontal resolution within the estuary and within the adjacent continental shelf.
In order to use model outputs, models should be scientifically sound and robust, and therefore must be validated to inform users how much confidence can be placed in the products. Model validation is a process that aims to demonstrate that a given site-specific model is capable of making sufficiently accurate simulations [
51]. Nevertheless, despite the importance of the validation process and although several statistical indexes have been developed and used for model evaluation [
51,
52,
53], there is a lack of a common validation procedure [
53]. Notwithstanding, the robustness of the models must be demonstrated which is only possible by comparing the model simulated results with in situ or remote observations such as satellite or HF radar products. Although the MOHID model is periodically validated, none of the works published until now focus exclusively on the validation of the 3D-MOHID for the TagusROFI. Only Vaz et al. [
47] validated the coastal area circulation model through the comparison of the superficial velocity predictions with HF Radar data for this domain in the vicinity of the estuary mouth for 2012. These authors found that model predictions do not fit perfectly with the HF radar data probably due to the high spatial variability of velocity that changes in spatial scales shorter than the model cell size. Nonetheless, in 3D hydrodynamic models, validation requires the comparison of the model results with at least sea level, current speed, and current direction [
54,
55], and since the model application is baroclinic, validation should also consider seawater temperature and salinity. Janeiro et al. [
30] validated a 3D application of the MOHID Water model for southern Portugal by comparing in situ measurements and HF Radar with predictions for all the five variables above. Although the TagusROFI domain has been applied in several research studies and as a tool for its management, this study will provide a higher insight concerning the validation of the results.
The present study aims to validate a 3D baroclinic application of MOHID Water model for the TagusROFI domain by using the above variables from a spatial-temporal perspective. With this purpose, model results will be compared with in situ measurements from several types of sensors (tidal gauges, CTD, and ADCP) and satellite L4 products (OSTIA, ODYSSEA, and MUR). The present work uses CTD, ADCP and satellite data to validate the TagusROFI domain for the first time. The main results are discussed, and conclusions are drawn concerning the model implementation in representing the coastal physical processes, its robustness to run in operational mode, including its forecast capability and consequently its ability to be a useful and powerful tool for coastal environmental management.
4. Conclusions
Regions of freshwater influence (ROFIs) are interface areas between terrestrial freshwaters sources and the ocean, where physical processes characteristic of both estuaries and of coastal areas often occur and overlap, making these areas very complex systems. The spatiotemporal scales of these systems are so large and complex that the collection of in situ and/or spatial data alone cannot explain the processes that characterize them. This can be overcome using numerical models that despite being an approximation of the reality, provide a continuous representation in time and space of the system variables. More than giving a better understanding of the interactions between the various components of the system, these models are essential for the monitoring and management of the costal systems and associated resources. Nevertheless, the robustness of model-derived products to address specific scientific and management issues depends on its accuracy. Models assume properties for which no measurements are available and have known inherent uncertainties (e.g., model simplifications, mesh generation, roughness definition, etc.) [
88]. Therefore, to minimize these uncertainties models are systematically calibrated by comparing modelling results with monitoring data. Apart from calibration, models should also be validated so as to determine the uncertainty associated with their solutions.
In the present work, the 3D-MOHID Water model was validated for the TagusROFI domain by comparing modelling results with in situ data (water level, surface seawater temperature (SST), salinity, and current direction and velocity) and SST remote sensing gridded L4 products (OSTIA, ODYSSEA, and MUR). These variables were chosen in the validation process because they are the most important hydrodynamics indicators with matching measured data. Indeed, the divergence of the horizontal flux (measurable through the propagation of the tide), in situ measurements of the velocity in water column, and the transport of conservative tracers (e.g., salinity) or of tracers with temporal variability lower than the tide (e.g., seawater temperature), are paramount parameters used to validate hydrodynamics. Seawater temperature and salinity are responsible for water density variability, which plays a central role in flow conditions in ROFI areas. Flow, however, can only be directly characterised with velocity fields and these can only be described through hydrodynamic models.
The statistical analyses used in validation include the calculation of the Pearson’s coefficient of correlation,
BIAS index and
RMES. Validation against two tidal gauges located at Cascais (coastal area) and at Lisbon (estuarine area) showed that the model is able to simulate observed water levels with accuracy, reproducing the amplitude and tidal phase, as well as its distortion during the propagation of the tidal wave within the estuary. Since the modelling results are accurate, it can be assumed that the model reproduces satisfactorily the horizontal flux divergence. A very good agreement between predictions of SST and salinity, and data from a CTD was also observed. The validation of current direction and velocity results using ADCP data indicated a high model accuracy for these variables. Finally, comparisons between model and satellite L4 products for SST showed that the model produces realistic SSTs and upwelling events. Although the use of more in-situ data would be desirable to increase spatial coverage, in the present paper it was used the most accurate data available. Notwithstanding, overall results showed that the 3D-MOHID Water setup and parametrisations were well implemented for the TagusROFI domain leading to highly accurate simulations. These results are even more important when a 3D model is used in simulations due to its complexity once it considers both horizontal and vertical discretization permitting a better representation of the heat and salinity fluxes in the water column. The validation of the 3D-MOHID Water is important because hydrodynamics is the basis for the transport of any biogeochemical variable through advection and/or diffusion processes, from the eulerian and langrangian point of view, and also indicates that it is robust enough to run in operational mode, including its forecast ability. Therefore, the 3D-MOHID can be considered an important management tool. Future research should be directed to the validation of the model’s biogeochemical component, and to the study of the biogeochemical processes in the TagusROFI domain. Moreover, it would be interesting to couple process-based-models (such as MOHID) with data-driven models [
89] that employs machine learning techniques. This approach consists in feeding data-driven models with data from process-based-models that were already validated. Although this hybrid approach is still in its infancy in what concerns hydrodynamics due to the complexity and variability of the inherent processes, it could be a useful approach in the coming future. The results of the present work constitute an important step towards the implementation of this innovative approach.