A high-resolution configuration of the Regional Ocean Modeling System model (ROMS; Shchepetkin and McWilliams, 2005) with embedded online two-way nesting domains capabilities and increasing resolution through Adaptive Grid Refinement In Fortran (AGRIF) package (Debreu et al., 2012) is used to simulate the circulation patterns at the Ría de Pontevedra during the year 2015. ROMS is a member of a general class of three-dimensional, split-explicit, free surface and a terrain-following s-coordinate numerical models, designed to resolve regional oceanic systems. ROMS solves the Reynolds-averaged Navier-Stokes equations using the Boussinesq and hydrostatic approximations, coupled with advection/diffusion equations for salinity and potential temperature, and the non-linear equation of state.

To investigate exchange processes between the Ría de Pontevedra and the coastal ocean, a multiple-nested grid with two-way nesting between first (FD), second (SD), and third (TD) domains was used (Figures 1A,B). An external large domain (LD) was implemented to properly solve the exchange between the Ría and the coastal ocean. The LD (Figure 1A) horizontal resolution is 1/27° (∼3 km in longitude), its boundary conditions are described in Nolasco et al. (2018). The LD domain also provides initial and boundary conditions, through offline nesting to the FD domain. The FD, SD, and TD domains (Figure 1B) have a horizontal resolution of 1/50° (∼1.6 km in longitude), 1/150° (∼0.550 km in longitude) and 1/450° (∼180 m in longitude), respectively, and are run with a timestep, dt, of 120 s for FD, 60 s for SD and 20 s for TD. SD was designed to solve inter-Ría and shelf circulation and TD solves simultaneously the three southernmost Galician Rias: Vigo, Pontevedra and Arousa (Figure 1C). Initial and boundary conditions for the SD, and TD domains are provided using two-way nesting with the FD and SD domains respectively.

The target domain, TD, has 30 sigma vertical levels employing stretching factors θ_b = 0 and θ_s = 5 to provide higher vertical resolution model outputs in the upper water column. The minimum depth (h_min) was set as 5 m. Bathymetry data were obtained from the RAIA observatory¹. Near the slope, smoothing was applied such that model bathymetry satisfied the condition that grid cell slope (r = dh/2h) did not exceed 0.2 (Haidvogel and Beckmann, 1999). A non-local, K-profile parameterization (KPP) boundary layer scheme (Large et al., 1994) is used to parameterize the unresolved physical vertical subgrid-scale processes. For lateral subgrid-scale mixing, the Smagorinsky model is used to define the horizontal viscosity and diffusivity, implemented through a rotated diffusion operator, as detailed in Marchesiello et al. (2009).

Several statistical estimators were used to compare observed and modeled data. The root mean square error (RMSE; Willmott, 1981) represents the sample standard deviation of the differences between modeled and measured values. The bias provides information about the model tendency to overestimate or underestimate the observed data, quantifying the systematic error of the model. The skill parameter (Willmott, 1981) is a measure of the quantitative agreement between the model and the observations. It considers the modeled and observed deviations around the observed mean, Xobs¯, to estimate the model representation, and varies between 0 (no agreement) and 1 (perfect agreement). These parameters are described as follow:

where X_mod and X_obs are the modeled and the measured data, respectively, and N is the number of analyzed data.

To examine the main modes of variability of the velocity fields, we perform an EOF analysis to these fields via singular value decomposition (SVD; Preisendorfer and Mobley, 1988). More details about the application of EOF methodology are provided in Supplementary Material.

Two different EOF analysis were applied to the hourly subtidal velocity fields, one at two horizontal layers: the surface and z = 20 m depth throughout the Ría de Pontevedra and another across the vertical cross-sections. The horizontal (vectorial) rvEOF analysis was computed for the meridional (V) and zonal (U) subtidal components of the surface velocity. The vertical (scalar) EOF analysis was applied to the normal component of the subtidal velocities obtained from the different cross-sections (Figure 2): the three openings at northern (SP6), central (SP5), and southern (SV2) cross-sections, and the middle (SP3) and inner (SP2 and SP1) cross-sections. Model output data down the water column was collected at vertical intervals of 1 m, every 300–400 m for across cross-sections SP1, SP2, SP4, SP5, SP6, and SV2, and every 182 m at SP3. The temporal PC from vertical and horizontal EOF analysis were compared with normalized meridional and zonal subtidal remote wind components.

Model accuracy was evaluated through statistical comparisons between field data obtained down the water column at eleven CTD-stations (Figure 2), a sea level gauge and model output, coinciding with field locations during the year 2015. Table 1 shows some statistical estimators (RMSE and skill) for the validation of total, tidal (astronomical) and residual sea surface height. RMSE is around 14–16 cm for total and tidal components and ∼8 cm for residual sea level. These RMSE scores for total and tidal components represent a small fraction of the mean tidal range (3.5 m). Skill scores are near one for total and tidal components, with exception of the residual component, 0.77. We can conclude that model output compares reasonably well with the sea level gauge data.

Table 2 summarizes the model accuracy in reproducing observed temperature (°C) and salinity data at eleven stations down the water column. Overall agreement between model and observed data is strong, with mean RMSE across the eleven stations 0.4 PSU and 0.8°C for temperature, mean bias 0.02 PSU and 0.3°C, and mean skill 0.9 and 0.95, for salinity and temperature, respectively. The highest disagreement with respect to salinity and temperature occurred at the most up-estuary site P0 (Table 2). While the lowest skill score was achieved for salinity at the most down-estuary site PA, located behind Ons Island (Figure 1C). Despite the differences between the model and measured data at some stations that are sheltered by the Ría topography, we can confirm that other nearby stations show good agreement.

Model results were compared with temperature and salinity data from two INTECMAR stations (Figure 2), for the year 2015. Comparison was done at 2 and 35 m depth for the outer Ría station, P4 (Figures 4A,C), and at 2 and 10 m depth for the inner Ría station, P3 (Figures 4B,D). Vertical salinity and temperature distributions were obtained at the outer Ría, P4 station (Figures 4E–H). Strong agreement between modeled and observed data was achieved throughout the seasonal cycle, at both depths (Figures 4A–D), as well as down the water column (Figures 4E–H). The model ability to replicate observed conditions is demonstrated by skill parameter values of 0.95 for salinity and temperature at P3 station, and 0.93 and 0.95 for salinity and temperature at P4 station. Temperature both near the surface (2 m) and at depth (10 and 35 m) (Figures 4A,B) shows nearly constant values of 12–13°C, from January to April. Variability in observed temperature and salinity during spring was well-reproduced, including temperature increases associated with the May downwelling event (Figures 4A,B). Reduction in salinity associated with the May 2015 downwelling/high rainfall event was well-reproduced at the outer Ría, P4 site. At the inner estuary site, P3, the model over-predicted the reduction in salinity in surface waters (Figure 4D, May). During summer and autumn months, the higher variability in surface and bottom temperatures is well represented by the model, including the reduction in temperature associated with the July 2015 upwelling event (Figures 4A,B,E,F). Vertical temperature stratification trends are also replicated, however bottom temperatures are typically over-predicted (higher than observed) (Figures 4A,B). From mid-November to the end of the year the characteristic cooling from late autumn to winter is also well reproduced by the model simulation (Figures 4A,B). Bottom salinity data (Figures 4C,D; red dashed line and dots) show little temporal variability throughout the year at both outer and inner Ría stations (Figures 4C,D) remaining at ∼35 and ∼35.5 PSU, respectively. Surface salinity is more variable, both temporally and spatially, with lower salinity (∼34.5 PSU at outer station and ∼33 PSU at inner station) associated with high river inflow. Surface salinity variability is well replicated by the model (Figures 4C,D,G,H). During the May 2015 downwelling event, a decrease in salinity is observed at both stations (Figures 4C–E,G) and at different depths. At the inner estuary, P3 station, the low salinity signal is due to high freshwater discharge from the Lérez river (Figure 3). At the outer Ría, P4 station, the low salinity signal is associated with water from the Ría de Vigo and the WIBP (Supplementary Video 1).

To investigate the effect upwelling has on circulation in the Ría de Pontevedra, a northerly wind event in July 2015 was investigated. This event was selected, as it commenced after the development of typical summer stratification of the water column, and after riverine runoff had decreased to summer baseline levels (∼5 m³/s). As a consequence, changes in temperature and salinity distributions could be attributed to upwelling. The upwelling-favorable wind event began on July 5 lasting until July 15, with 5–10 m/s northerly winds blowing from July 5 to 9 and light southerly winds blowing on July 10, followed by ∼10 m/s northerly winds blowing from July 11 to 15. The effect of upwelling on the Ría circulation was investigated by analyzing model 3D current patterns extracted at 13:00 h, July 12, 2015 (Figure 5), at seven cross-estuary transects and at the outer Ría, P4 station (Figure 2). This day was selected to visualize Ría conditions, since on that day the upwelling signal had reached central and inner Ría sites. Daily mean currents were analyzed to determine the subtidal estuarine circulation patterns associated with the upwelling event, the results of which are plotted in Figure 6.

Modeled subtidal circulation at the three openings and at the middle and inner Ría revealed a two-layer circulation whereby surface currents flowed down the Ría, while near the bed currents flowed into the Ría (Figure 5; SP1, SP2, SP3, SP5). Between the outward, and inward flowing currents, a level of null currents occurs. The null level was located at depths of ∼20 and ∼10 m for SP5 and SP3, respectively, and at ∼5 m for both SP2 and SP1. Flow in/out of the estuary differed at the three different Ría openings. For the main Ría opening (SP5) outward flow was directed S/SW, driven by Ekman flow, while at depth inward flow was directed E/NE, aligning with the orientation of the main Ría channel. At both SP6 and SV2, which are shallower, narrower entrances, currents at the surface and at depth were dominated by the presence of the equatorward coastal upwelling jet, with currents flowing into the estuary via SP6, the northern opening, and out of the Ría, via SV2, to the south (Figure 5). In the central and inner Ría (SP3, SP2, and SP1), the direction of inward and outward currents appears to be influenced by both Ekman flow, and topographic steering, since the currents align with the estuary axis at these sites.

The typical estuarine pattern described above (Figure 5) with outflow (inflow) through the surface (deep) layers, was analyzed in more detail through the daily mean along-Ría section (Figure 2) over successive days. The time evolution of temperature, salinity, density and subtidal horizontal and vertical velocity fields, from July 11 to 15, 2015, are shown in Figure 6. Conditions inside the Ría indicate maintenance of the upwelling event over the five days investigated. On July 11 (Figures 6A,B), the river plume is confined to the inner part of the Ría, where salinity is relatively low (<35 PSU). In the inner Ría, horizontal velocities are low (<6 cm/s). Surface wind forcing, and the baroclinic pressure gradient induced by the low river flow, generates offshore transport in surface waters. At the same time, colder shelf waters flow into the Ría via the main opening at depth (below 20 m). This bottom inflow of colder shelf waters is observed by the rising of the isohaline 35.65 PSU (Figure 6A) and the isotherm 14°C (Figure 6B). On July 12 (Figures 6C,D), maximum wind velocities reached 10.9 m/s from the northeast. The setup of the Ekman flow at the surface in the outer Ría and adjacent shelf favors the offshore transport of the Lérez river plume. Inshore flow at depth is well established, as demonstrated by the upwelling of the isolines which rise up to 10 m/day. This corresponds with an estimated vertical velocity of 0.11 mm/s, which is comparable to the modeled velocities. Results from July 13 demonstrate increasing stratification between outward flowing, warmer surface waters and colder shelf waters flowing into the Ría at depth. The continued flow of cold water into the Ría causes isohalines to rise, confining the fresher, warmer water to the upper few meters (<10 m) of water column. On July 14, stratification continues, however velocities at the outer estuary decrease down the water column. By July 15 (Figures 6I,J), stratification in the outer estuary decreases, with surface water temperatures reducing relative to previous days. In the inner Ría, over successive days, bottom water temperatures reduced, and stratification with respect to temperature and salinity increased over successive days.

A vigorous spring downwelling event occurred from April 30 to May 4, 2015 in response to strong (∼10 m/s) southerly winds (Figure 3B). The downwelling-favorable winds coincided with high river runoff from the Lérez River (maximum 84 m³/s on May 4, 2015). In the days preceding the downwelling event, winds were variable, blowing from the northeast and south, with southerly winds coinciding with higher river flow (Figure 3).

Velocities at cross-sections during the downwelling event are depicted in Figure 7 for April 30 at 13 h, the first day of the event. As with the previously described upwelling event, two-layer circulation developed along the main Ría axis, however circulation was reversed, with surface currents flowing up-estuary despite high river flow, and currents at depth flowing down-estuary. At the southern entrance (SV2), currents flow into the Ría at all depths while at the northern opening (SP6) currents flow out of the estuary at all depths (Figure 7). The null level, where residual currents were essentially zero, were located between 30 and 40 m at SP5, ∼20 m at SP3 and at ∼5 m for both SP2 and SP1. The surface inflow during this downwelling events has a prevailing NE direction associated with the poleward coastal downwelling jet, whilst the bottom outflow has a W/SW direction.

Along the main Ría axis a well-established inverse estuarine circulation, characterized by surface up-estuary flow and down-estuary flow at depth, is evident from model results, replicating conditions on April 30, 2015 (Figures 8A,B). Lower-salinity waters associated with high river discharge from the Lérez are confined to the inner Ría, where salinity reduced to 30 PSU. On May 1 (Figures 8C,D), two vertical clockwise (CW) cells developed at the outer and middle Ría, with a negative estuarine circulation. Also, a transient counterclockwise (CCW) vertical cell, located below the river plume, establish a positive estuarine circulation at the inner Ría. In the outer Ría, the near bottom layer outflow generates ∼20 m downwelling of the 15.5°C and 35 PSU isolines (Figures 8C,D). This downwelling corresponds with an estimated vertical velocity of −0.23 mm/s, which is comparable to the modeled velocities. On May 2, the vertical cell at the outer Ría was displaced toward the head of the Ría, originating a mixing of the water column at the middle Ría. On May 3 (Figures 8G,H), strong vertical mixing continues, with vertical velocities exceeding 0.1 mm/s. σ_t isoline distributions follow the distribution of isotherms at the beginning of the event, but as river discharge increases they more closely follow isohalines. During the last day of the spring downwelling event, May 4 (Figures 8I,J), southerly wind speeds reached ∼12 m/s causing the river plume to spread offshore in the direction of the northern mouth (SP6; see Supplementary Video 1 for further details). Salinity stratification propagated to the outer Ría over time, with the strongest degree of stratification evident on the last day of the event (Figure 8I).

A rvEOF analysis at 0 and 20 m depths (horizontal EOF analysis; Figure 9), representative of a surface layer and deeper layer (usually surface and deep layers have opposite circulation patterns) was done, providing a synoptic overview. Figure 9 represents the spatial fields of the horizontal rvEOF analysis, starting with the average spatial fields (Figures 9A,B) at z = 0 and z = −20 m, followed by the three first rvEOF spatial modes, rvEOF1 (Figures 9C,D), rvEOF2 (Figures 9E,F), and rvEOF3 (Figures 9G,H), accounting for 66.1, 13.3, and 4.7% of the variance, respectively. Visualization of the temporal variability of the horizontal fields rvPCs series are shown in Supplementary Figure 1.

An EOF analysis was also applied to subtidal velocities normal to the cross-sections SP6, SP5, SV2, SP3, SP2, and SP1 (vertical EOF analysis; Figure 10). Vertical EOF analysis, depicted in Figure 10, represents the EOF modes through the temporal principal component, PC1 (Figure 10A) and PC2 (Figure 10B), the average spatial fields (Figure 10C), and the spatial EOF1 (Figure 10D) and EOF2 (Figure 10E) that account for 60.8 and 12.5% of the variance, respectively.

The Ría de Pontevedra acts like a partially mixed estuary with a two-layer circulation (Klinck et al., 1981) throughout the year, driven principally by Ekman dynamics over gravitational circulation. This circulation pattern presents a null current level between these two layers, whose depth varies along the Ría. Gómez-Gesteira et al. (2001) observed and studied the displacement of this null level at the Ría de Pontevedra, and the null level they reported in the inner Ría are like those predicted by our model. At the outer central opening (SP5), Aguiar Fernandez (2016) located this null level at same depth (20 m depth) as we do. This depth can vary depending on wind intensities. For example, Barton et al. (2015) observed this level at a shallower depth, 15 m, at the outer Ría de Vigo. The Ría de Pontevedra has three openings that exchange waters with open ocean (denoted SP6, SP5, and SV2 in Figure 2), however it differs with neighboring Rías, Ría de Vigo and Ría de Arousa, which have two openings. The southern opening (SV2) of the Ría de Pontevedra coincides with northern opening of the Ría de Vigo. At these entrances, outer Ría circulation, behaves as a continuation of offshore shelf circulation (Largier, 2020) during upwelling and downwelling events, whilst at the middle and inner Ría, circulation patterns are driven by buoyancy gradients, local wind forcing, river inflow, and topographic steering (Álvarez-Salgado et al., 2000; Barton et al., 2015, 2016; Gilcoto et al., 2017). Shallower openings located at the North/South of the Ría de Pontevedra (SP6 and SV2), respond with a one-layer circulation, whilst the central opening (SP5) is deeper than the Ekman depth, and presents the characteristic two-layer upwelling/downwelling circulation (Souto et al., 2003; Gilcoto et al., 2007). This behavior is schematically summarized in Figure 11.

Concerning the alongshore transport in the inner shelf region, we remark that at shallower entrances, the coastal jet flows in through the northern (SP6) opening, and leaves through the southern (SV2) opening under northerly winds, and vice versa under southerly winds (detailed information about the circulation patterns are shown through video-animations of surface and deep circulation, as well as wind, available at Supplementary Videos 1, 2). Coastal jets are not generated under zonal wind components (Csanady, 1982). In the Rias Baixas, the coastal jet, driven by the alongshore wind component has elements of an upwelling shadow and upwelling trap, where the coastal jet splits flowing either side of the outer Ría islands (Barton et al., 2015; Aguiar Fernandez, 2016; Piedracoba et al., 2016; Largier, 2020). Coastal jets are described by Largier (2020) as one of the three primary drivers of circulation in upwelling bays, together with local wind stress, and bay-ocean differences in buoyancy and water level. As observed in Supplementary Videos 1, 2, under downwelling conditions coastal jets may also transport rivers plumes from neighbor rivers. This process contributes to the decrease of the Ría salinity and has an influence on the Ría density gradients, which reverse the circulation, allowing a rapid renewal of estuary waters (Barton et al., 2015; Largier, 2020). In addition, the presence of river plumes and its advection by the coastal jet increases the exchange of properties between the Ría and the shelf. This assumption needs further analysis through Lagrangian studies.

The narrow width and the shallower depths of the inner Ría, induces a response in the water column to the wind forcing, with a direct downwind transport in the surface layer, and an opposite motion at the bottom layer. This response is likely induced by along-Ría axis (approximately zonal) winds, as we described through its correlations with the second EOF mode. Westerly/easterly winds force surface water to flow in/out of the Ría, and hence a bottom outflow/inflow. Consequence of these opposite two-layer flow, the vertical stratification increases, reinforcing the ocean-Ría exchange (Gilcoto et al., 2017; Largier, 2020). de Castro et al. (2000) provide further evidence of this two-layer response, driven by zonal winds at the Ría de Pontevedra during April 1998. Also, in the Supplementary Videos, we can observed bilayer circulation patterns consequence of westerly winds from January 29 to 30 (Supplementary Video 1), August 13 and September 13 (Supplementary Video 2), and consequence of easterly winds from February 9 to 10 (Supplementary Video 1), October 18 and November 2 (Supplementary Video 2), for instance.

As a consequence of the increase of Lérez river inflow, river plumes spread along the northern coast (associated with Coriolis forces), after exiting the Ría through the northern opening (SP6; Figures 9A,E). Similar behavior has been observed at the Ría de Arousa (Rosón et al., 1997) and at the Ría de Vigo (Sousa et al., 2014b), whose plume penetrates into the Ría de Pontevedra through SV2 opening and leaves through the central opening (SP5; Figure 9A). These circulation patterns are in agreement with the observations discussed by Aguiar Fernandez (2016), based on ADCP and CTD surveys. Some episodic events, also modeled with our configuration (not shown), show intrusions of river plumes from the Douro and Minho rivers (WIBP), which originated south of the Ría de Vigo, as reported by Des et al. (2019) and Mendes et al. (2016).

A vertical flow may be enhanced during downwelling events, whilst upwelling events are characterized by a stratification of the water column (Largier, 2020). Evidence of these processes are illustrated by an elevation/sinking of the isohalines about 10–20 m per day at the outer Ría during the upwelling/downwelling events (Figures 6, 8). Pardo et al. (2001) observed the rising of the 35.5 PSU isohaline at the Ría de Pontevedra during a river flow increase and a relaxation of southerly winds, but they did not quantify it. At the Ría de Vigo, Barton et al. (2016) reported a sinking of isopleths, at inner and middle Ría stations, of 8 and 12 m/day, respectively, which are similar to the results reported in this study at the outer Ría de Pontevedra. This displacement, during downwelling events, generates a blocking of the horizontal circulation at the middle Ría, due to a front (Figures 8C,D,G,H) splitting two clockwise vertical circulation cells. We hypothesize that these two clockwise circulation cells or negative estuarine circulation, are closely associated to the presence of the different basins at the Ría de Pontevedra. One cell between SP5 and SP3 sections, is associated to the outer basin, and the other, located between SP3 and SP2 section, is associated to the middle basin. Evidence of a third vertical cell, with a counterclockwise rotation (Figure 8D), is located at the innermost basin of the Ría, associated to the influence of a high riverine flow (Figure 3A; blue box), that induces a positive estuarine circulation at this location. A similar distribution of vertical circulation patterns is discussed by Villacieros-Robineau et al. (2013) with two opposed vertical circulation cells during downwelling events with high river runoff, at the Ría de Vigo. On the other hand, Barton et al. (2016) refer that the convergence flow and the time evolution of the vertical circulation cells during downwelling events are strongly controlled by the wind-driven surface inflow from outside of the Ría and the strengths of accumulated freshwater outflow at the inner Ría.

Influence of local and remote winds have been thoroughly discussed by Gilcoto et al. (2017). These winds are highly correlated, leading to difficulties in the interpretation of the dynamic response of the induced circulation. However, they found that remote winds better explain subtidal residual circulation, relative to local winds. Regarding influences of local and remote winds, we note that our wind point is located at an intermediate location between the remote and local wind points used by Gilcoto et al. (2017), but since the resolution of WRF model run by MeteoGalicia is 4 km, remote winds are better captured by our model configuration.

Closed circulation features (Figures 9G,H), displayed in spatial third EOF mode are usually associated to recirculation processes, and consisting of cyclonic or anticyclonic vorticity structures. They are characterized as transient, with timescales of one or two days. The occurrence of these vorticity transient structures is mainly related to abrupt changes in wind intensity and direction. At the Supplementary Video 1 (i.e., February 16, April 12, June 15) and Supplementary Video 1 (July 3, September 18, October 8) we can observed these circulation features. At the Ría de Vigo, these structures have been thoroughly described by Piedracoba et al. (2016). Also, Barton et al. (2015) describe them as consisting in a cyclonic eddy behind Cíes islands during a downwelling event, and near the middle Ría during an upwelling event. In addition, Piedracoba et al. (2016) observed a dipolar vorticity structure in the Ría de Vigo, with positive vorticity in the northern outer region and negative vorticity in the southern region during downwelling conditions. During upwelling conditions, behavior reversed. Differences in the intensities of the vorticity structures, were shown to be related to variable strength of upwelling and downwelling events (Piedracoba et al., 2016).

Several circulation features (coastal jets, fronts, vertical cells, horizontal closed circulation, and circulation blocking) discussed above, are associated to exchange of water properties between the shelf and Ría (including biogeochemical properties and pollutants, among others), and/or retention of biological material (nutrients and phytoplankton, for instance).

The Ría de Pontevedra is strongly affected by Harmful Algal Blooms events, which lead to closure of raft mussel farms (227 ± 57 days/year during 1998–2006 according to Álvarez-Salgado et al., 2008). Closures cause serious financial difficulties for the harvesting mariculture industry and represent a challenge for the sustainability of mussel farming resources. As such, having tools capable of forecasting HABs events in the Rias Baixas, both under present and future conditions, would prove vitally important to the management of marine resources. Although the present study is not specifically focused on HABs forecast, the improved understanding of circulation processes presented in this study provide clues to better understand the physical mechanisms associated with its development. We applied EOF techniques to subtidal Ría velocities, which established a good correlation between the meridional and zonal wind components (main forcing of the circulation) and the first two EOF modes of the horizontal velocities, respectively. Thus, high-resolution modeling and statistical analysis, such as EOF, allow us to predict the ocean-Ría exchanges as well as the circulation inside the Ría, when wind forecasts are available. Future research will implement the findings from this study, along with additional model simulations to determine the physical Ría conditions that allow Harmful Algal Blooms to develop.