The Sagres peninsula and Cape Saint Vincent are located in the extreme south-west of the Iberian peninsula (Figure 1). The coastal area has minimal input of terrestrial nutrients due to a low human population of circa 7,000 inhabitants (Loureiro et al., 2011), geographic position and an absence of significant terrestrial fluvial input. Insignificance of the terrestrial sources suggests the main source of nutrient enrichment of the coastal waters in the study area is wind-driven upwelling from nutrient-rich deep waters. The sampling station was located approximately 5 km east from Cape Saint Vincent, and 1 km south of Sagres port (37°00′39″ N, 8°53′58″ W), near the offshore long-line installations for bivalve aquaculture (Figure 1), and has a depth of 25 m. In total, 266 samples were obtained, over the 8-year period starting from 27th January 1994 until 16th October 2001. Periodicity of the sampling was not regular, if sea conditions permitted, usually bi-weekly, with a 45 μm mesh size plankton net to target dinoflagellates of medium and large size; the net was towed horizontally for 5 min in the surface water layer at 1 m depth; sampling took place during late mornings at 11–12 a.m. to minimize the effect of diurnal vertical migration. The filtration method was chosen to focus on the collection of armored dinoflagellates that were usually present in low densities and to increase the probability of detecting rare species, simultaneously some unarmored species larger than the mesh size (45 μm) were also collected. Samples were fixed with 1% formaldehyde and stored refrigerated (+8°C) before analysis. Microscopic analysis was performed using 100× to 400× magnifications. Approximately one hundred (on average 106, min. 78, max. 185) random cells were identified to species level from each sample. Taxa identification was performed using Dodge (1982) and Steidinger and Tangen (1996) as sources of taxonomic information. Therefore, the dinoflagellate species abundance data was proportional, reflecting relative abundances of dinoflagellate species compared to the total number of the individuals identified in each sample, which was in most cases close to one hundred cells per sample.

Upwelling conditions were characterized based on upwelling indices derived from wind stress and sea surface temperature (SST) to study the influence of upwelling on the phytoplankton community.

Sea surface temperature (SST) time series were obtained from the NOAA “SST Daily Optimum Interpolation (OI), AVHRR Only, Version 2.1, Final, Global, 0.25°, 1981-present” product (Reynolds et al., 2007), for the data grid square containing sampling station, with coordinates range N 36.75° to N 37.00°, W 8.75° to W 9.00°. Wind speed and direction components were obtained from the “Blended Daily Averaged 0.25-degree Sea Surface Winds product” (at 10 m level), made available by the National Oceanic and Atmospheric Administration (NOAA) and National Climatic Data Center (Zhang et al., 2006), from the grid square centered at N 37.00°, W 9.00°. The wind conditions at that point were considered representative of the local upwelling, although the study area belongs to the broader upwelling system, influenced by winds along the coastal stretch of Western Iberia. It was previously shown (Gonzalez-Nuevo et al., 2014) that upwelling indices based on local data were a good approximation for regional conditions for the NW Iberian upwelling system.

Wind stress upwelling index (UI) was based on Ekman transport (Q_x,y) of surface water calculated following Bakun, 1973 and Cropper et al. (2014), as

where τ_x,_y is the wind stress, u and v are the wind speeds in longitudinal and meridional directions respectively, w is the module of wind speed, ρ_a is the air density (1.22 kg m^–3), C_d is the dimensionless drag coefficient (1.14 × 10³) (Large and Pond, 1982), f is the Coriolis parameter (8.78 × 10^–5 s^–1 for the study area) and ρ_w is the density of sea water (1,025 kg m^–3).

The Ekman transport latitudinal (Q_x) and longitudinal (Q_y) components were used to represent the upwelling indices for the western and southern coasts, because the coastline directions in the study area were roughly parallel to the meridian and the equator. Negative values of Q_x and Q_y indicated upwelling favorable conditions along the western and southern coasts in the study area, respectively. The latitudinal upwelling component of the western coast (Q_x) is more important in Sagres area (Krug et al., 2017), as its influence spreads around the CSV. Cumulative upwelling index (CUI) was used to represent summation of the daily upwelling indices (Q_x) from January 1st to the end of every year, and contributed to the estimation of the length of upwelling season and general upwelling intensity through the years [as in Díaz et al. (2016)].

Surface Chl a was used as a proxy of total phytoplankton independent of the dinoflagellates abundance, based on the results of previous studies on phytoplankton pigments in the same area (Goela et al., 2014). SeaWiFS Chl a product was available for the dates from 15th September 1997 until the end of the study period in 2001. Data on Chl a was obtained via the NOAA Environmental Research Division Data Access Program (ERDDAP) via web-site http://upwell.pfeg.noaa.gov/erddap/index.html.

Statistica 6.0 was used for correlation analysis and basic statistical tests. Principal component analysis (PCA) was used as an exploration method to reveal the structure of species composition. The set of species used for this analysis was selected by their abundance across samples, in order to focus on bloom forming species. Usually abundant species tended to be highly frequent as well, so this approach did not distort the representation of common species. As PCA is sensitive to values scale, abundant species would account for greater explanation of the variance. The data was normalized (across samples) and standardized (by subtracting the mean and dividing by standard deviation) before analysis to account for both the higher and the lesser abundant species. PRIMER-E version 6.1.6 (Clarke and Warwick, 2001) was used for calculations of PCA and Similarity Percentages of species contributions (SIMPER) analysis and data transformations.

The total species richness of dinoflagellates during the 8-year study period comprised 194 species, across the 266 samples examined. Taxa names with the respective short codes and frequency of occurrence are listed in the Supplementary Appendix 1. Sixty most common species occurred in not less than 26 samples (10% frequency) and 41 species in not less than 53 samples (20% occurrence frequency) out of the total 266 samples examined (Table 1). The most frequently identified species belonged to genera Tripos, Protoperidinium, Dinophysis, Diplopsalopsis, and Prorocentrum. On average 25.5 species per sample were identified, the average Shannon community diversity index was 2.53 and ranged from 0.34 to 3.40. Only 107 species occurred in quantity of more than 10 individuals and 62 species of more than 50 individuals throughout the study period, indicating that the sampling method was effective for detecting rare species.

The most frequently occurring potential HAB species by order of occurrence were Dinophysis caudata, Phalacroma rotundatum, Gonyaulax digitale, G. polygramma, Dinophysis acuta, Lingulodinium polyerda, Gonyaulax grindleyi, G. spinifera, Alexandrium affine, D. acuminata, D. ovum, Prorocentrum gracile, and P. micans.

Principal component analysis (PCA) was chosen to study community structure, as it was considered to be the appropriate method for the analysis of proportional datasets, i.e., lacking density values and describing community in terms of relative proportions of species. The PCA analysis was performed on standardized and log transformed abundances of the 50 most frequent species (Figure 5), to reduce the influence of the most abundant species and reveal the variability associated with relatively rare species.

Sampling dates were separated by the seasonal factor, summer and autumn samples were distributed mainly below the center of diagram (negative PC2 scores). Winter and spring samples were associated with positive scores on PC2, but were widely scattered along the PC1, probably because of the influence of the most frequent species. Group 1 (as in Table 3), consisting mostly of Tripos spp., was clearly associated with sampling dates from November through April-May. Group 2 (Protoperidinium spp.) was distinctly separated and associated with spring (February to May) sampling dates. Vectors of species belonging to Groups 3 and 4 were more weakly separated, but linked with summer and autumn sampling dates. These two groups contained most of the HAB species typical of the area, belonging to genera Dinophysis, Prorocentrum, Gonyaulax, and Lingulodinium polyedra.

Taking into consideration the low total variance explained by the first three PCs (20.76%); a consequence of a large number of variables (species), reflecting the complexity of community. Therefore, the results for the PCA are treated as descriptive, but four types of dinoflagellate assemblages were separated by the combined effect of the variances associated with the first two principal components (Figure 5).

In order to further reveal seasonal differences, PCA analysis was applied to species proportions averaged by season of each year. The resulting plot demonstrated seasonal differences in the composition of dinoflagellate community (Figure 6). Winter and spring communities were distinctly separated from summer and autumn, the latter two being more similar. Comparison of the species variables vectors with data points in the principal components space demonstrated that species groups responsible for seasonal differences were similar to groups identified in the previous PCA analysis done on the non-averaged data. Winter communities were mainly characterized by the Tripos group (T. furca, fusus, horridum) and others, Triadinium polyedra, and the Protoperidinium group (P. leonis, mite). In spring, the more numerous of the Protoperidinium group (P. longipes, P. depressum) and others were highly represented, together with the Diplopsalis group (incl. Preperidinium meunieri) and the constantly present Tripos genus. Summer and autumn communities showed less distinct separation, with main contribution from the presence of Protoperidinium divergens, Dinophysis caudata, Lingulodinium polyedra and Prorocentrum triestinum. Summer and autumn were more likely to contain HAB species.

Similarity percentages analysis (SIMPER) was further used to determine the contributions of individual species to the similarities between samples within different seasonal groups (Table 4), based on the Bray-Curtis similarities of the resemblance matrix. In each case, the average similarity between samples belonging to seasonal groups was in the range from 42 to 46%, with 12–16 most significant species responsible for 75% of total similarity. Generally, the most frequent and relatively abundant species contributed the most to similarity in all seasonal groups. In winter, Tripos furca, fusus, horridum and muelleri, Dinophysis caudata and several Protoperidinium spp. were the main contributing species. In spring samples, similarity was driven by the same Tripos species with a higher contribution of Protoperidinium depressum, longipes, divergens, and ovatum. In summer, Dinophysis caudata contributed to similarities, together with the same frequent Tripos and Protoperidinium species. Autumn samples were characterized by decrease of contribution of Protoperidinium, with Tripos furca, T. fusus, T. muelleri and Dinophysis caudata being the most important contributors. In general, similarities between communities of the same season revealed by SIMPER were attributed to the most common and abundant species, but not specifically to HAB species. The exception was Dinophysis caudata, which was the most common harmful species that contributed to seasonal dissimilarity.

The species assemblages described in this study were defined as the result of observations of coincident occurrence, and the study did not examine any inter-species interactions; therefore, we did not consider here the existence or influence of such interactions on the community composition. The data collected was limited only to larger dinoflagellates and therefore, other primary producers, such as other phytoflagellate classes and diatoms, were not included. Considering this, the composition of the dinoflagellate groups is treated here as a contingent abundance of species, that are promoted by favorable oceanographic conditions. Predation and other ecological interactions could not be assessed because of the limitations with the method.

Taxonomically closely related species, i.e., belonging to the same genera tended to co-occur together, as revealed by multivariate analysis. This likely reflects similarity in their ecological requirements. This relation in occurrence was especially pronounced in common genera such as Tripos, Dinophysis and Protoperidinium. A high similarity was observed between Tripos furca, T. fusus, T. horridum and T. muelleri that occurred in association with Protoperidinium longipes, P. divergens and Dinophysis caudata, meaning that such frequent species tended to have similar occurrence and abundance patterns, therefore, were favored by the same environmental conditions.

Observed grouping of the taxa in this study was similar to dinoflagellate life-form types delimited by Smayda and Reynolds (2001, 2003). In the present study, the “Tripos” community (Group 1) seemed to be associated with winter and spring samples, so outside of the upwelling period, but in some cases also autumn (Figure 5). We found Group 1 of this research (“Tripos”) to be comparable with Type III by Smayda and Reynolds (2003). This dinoflagellate type was described by the same authors to bloom during summer and autumn stratification in coastal waters, but also favored by elevated nutrients concentrations, which was partly opposite to our observations. This could be explained by the prevalence of upwelling conditions during summer, while less active upwelling combined with relatively high SST during Mediterranean winter were possibly favorable for the “Tripos” group of species.

Group 2 was very distinct and contained almost exclusively Protoperidinium and taxonomically related genera Diplopsalopsis, Oblea, Preperidinium with Protoperidinium depressum, P. simulum, P. ovatum, P. leonis, P. conicum and Diplopsalopsis bomba occurring as the most abundant species. It was usually associated with spring and summer sampling dates but occasionally found in autumn. It can be compared with Type II (nutrient enriched warm conditions) in Smayda and Reynolds (2003). Data obtained at Sagres area during 2010–2012 (Goela et al., 2014) suggests that high Protoperidinium and Diplopsalis-type abundance was associated with spring-summer blooms characterized by high biomass chain forming diatom development (Guinardia spp., Chaetoceros spp.) and was related with upwelling conditions favoring development of diatoms (nutrient enrichment, deep mixed layer, turbulence).

Group 3 contained such HAB species as Dinophysis caudata and Lingulodinium polyedra together with Prorocentrum micans and P. scutellum. These species can be compared with Type II assemblage described by Smayda and Reynolds (2001, 2003) that are formed during summers in conditions of rather elevated nutrient levels and local stratification. In this study, Group 3 was clearly associated with the end of summer upwelling season and early autumn (July—November), and so corresponded with Type II habitat conditions. Type V (Upwelling relaxation taxa) also can be compared to Group 3 (incl. by the presence of L. polyedra). Therefore, Group 3 species combined features described for Types II and V assemblages and could be associated with the conditions of establishing stratification at the end of the upwelling season during summer and autumn.

Group 4 was the most taxonomically diverse and contained Protoperidinium longipes and Dinophysis acuta, as well as other Dinophysis species (D. fortii, D. acuminata, D. ovum), Gonyaulax spinifera, Scrippsiella trochoidea, Pyrophacus sp., Torodinium sp., Ornithocercus sp. and Noctiluca scintillans. It was typically associated with spring and summer samples and by taxa composition possessing the traits of Type VII (Dinophysoids) and VIII (Tropical taxa) described by Smayda and Reynolds. These types represent habitat range from small scale upwelling currents to oligotrophic stratified and usually offshore conditions.

Type I (Gymnodiniods) by Smayda and Reynolds (2001) could not be compared with the results of this study because only larger armored dinoflagellates were collected. However, in the data obtained at the same location during 2010–2016 (Goela et al., 2014; Danchenko et al., 2019) Gymnodinioid species were frequently present during summer upwelling season, often concurrently with diatom blooms.

The present study does not allow direct comparison of dinoflagellate community development with that of the other phytoplankton groups (diatoms, flagellates, coccolithophorids). Described patterns of seasonal blooms of diatoms, nanoflagellates and cyanobacteria (Loureiro et al., 2005, 2011; Goela et al., 2014) in the same area indicated that late autumn and winter (November to February) are the seasons with generally low abundance of phytoplankton, dominated by nanoflagellates. The corresponding winter dinoflagellate community in this study contained species of the “Tripos” Group 1 and some rare species in low proportions. Spring and early summer at study site were characterized (Loureiro et al., 2005, 2011; Goela et al., 2014, 2015; Danchenko et al., 2019) by the proliferation of blooms, mainly composed of diatoms belonging to genera Chaetoceros, Guinardia, Pseudo-nitzschia and marked by an increase in Chl a. These spring blooms could be initiated by the upwelling-driven increase in nutrients, increasing photoperiod and SST (Loureiro et al., 2005, 2011; Goela et al., 2014). In the current study, the dinoflagellate community that developed during spring was characterized by Group 2 (“Protoperidinium”). June and August were normally the periods of active upwelling associated with N and W winds, which decreased by the end of September—October (Haynes et al., 1993; Goela et al., 2016). Diatom blooms are known to decline in such conditions and replaced by community dominated by nanoflagellates and by warm oligotrophic water taxa of dinoflagellates that we described as Group 4 (Dinophysis acuta, D. fortii, D. acuminata, D. ovum and Gonyaulax spp.). Groups 3 and 4 contained a majority of potentially HAB species. Therefore, upwelling relaxation in summer and autumn is the season when conditions for harmful blooms of dinoflagellates can be expected in the study area.

It was revealed by both PCA and SIMPER analyses that the most common and frequently occurring species made the main contribution to statistical measures of similarities between samples of the same season and similar oceanographic conditions, as well as to dissimilarities between different seasons. Harmful species, except for Dinophysis spp., were not found to be among the most abundant and frequently occurring (Table 5). Other factors affecting communities (such as ecological, competition, life-cycle and others) were not examined in this study, but possibly contributed to the high temporal variability observed.

One of the objectives was to provide information on harmful algae monitoring in the interests of shellfish aquaculture activities. The observed co-occurrence between HAB-containing types of assemblages with those of the more common species suggest the requirement for similar ecological conditions. The communities with harmful species had a general tendency to develop during the relaxation conditions of late summer-autumn, that followed a long summer upwelling season. There were nevertheless cases of HAB containing assemblages in winter (e.g., Dinophysis spp. in spring 1995, 1996 and notably from summer 1997 through winter until spring 1998, that corresponded with DSP event; L. polyedra present since March to August 1997). Dinophysis species, especially D. acuminata were often detected in low numbers throughout the annual cycle, allowing them to increase in any season under favorable conditions.

In the West Iberian region (Galicia, West Portugal), the main source of dinoflagellate-related HABs are Dinophysis species (Escalera et al., 2010; Díaz et al., 2016; Moita et al., 2016). Early onset of upwelling and upwelling events during spring were reported to promote D. acuminata blooms in Galician Rias (Díaz et al., 2013). Hot summers with positive SST anomalies, with weak upwelling conditions were found to be associated with the early decline of population of D. acuminata and its exchange to Dinophysis acuta (Díaz et al., 2016, 2019). During this study, the most pronounced Dinophysis contribution was observed starting summer 1997, and lasted through the winter, and during February to September 1998 coincided with DSP outbreak in the Algarve, reported by Vale et al., 1999. Dinophysis bloom in February-March 1998 was represented mostly by D. acuminata and to lesser degree by D. caudata, later in summer D. acuta appeared and reached high numbers (together with D. caudata) in July to October 1998. Such pattern is similar with reports from N Iberia by Díaz et al. (2013). D. acuta was also significantly represented in other years, e.g., 1994. Study sampling site at Sagres was located to the West from Ria de Alvor and Ria Formosa, where the DSP health event was observed, therefore a direct comparison is not feasible. But nevertheless, the coincidence in time and lengths with the recorded outbreak suggested that similar conditions for Dinophysis blooms existed along the South Portugal coast. Coastal counter-currents directed East to West were found to be independent of the local wind forcing, defined by regional barometric features and frequent (Sánchez et al., 2006; Garel et al., 2016) along the S Algarve coastline. These warmer counter-currents may have contributed to warmer and more stratified conditions at the shelf, contributing to the DSP event in 1998 (Vale et al., 1999), and the increase in Lingulodinium polyedra observed in some years. The influence of Gulf of Cadiz counter-current is likely to be more pronounced in the Eastern Algarve coast, compared to the sampling site at Sagres near Cape San Vincent.

The regular shellfish quality control required by law is implemented by IPMA (Instituto Português do Mar e da Atmosfera) since 1986, and includes sampling of phytoplankton and commercial bivalves (Vale et al., 2008; Anonymous, 2013). The reports were made public since 2014, and so do not cover the period of this study. The suggestion resulting from this study would be to manage the timing of collection of shellfish so that it does not coincide with upwelling relaxation in summer-autumn, which can be assessed from the published meteorological data. In the Sagres and Cape Saint Vincent area this means relaxation of prevailing N and NW winds and change to winds of other directions during late summer and autumn. Adjustment of harvesting based on the knowledge of weather and oceanographic conditions may assist the bivalve aquaculture stakeholders to reduce the probability of toxicity in shellfish and compliance with safety requirements.

Long term studies of trends in climatological wind patterns (Leitão et al., 2018) indicated that the annual Northerly wind component has tendency to increase along the coasts of Iberia, including South and South-West coasts over the recent decades. At the same time, no changes of wind trends were found for summer, and decrease of N winds was observed for September. With regards to upwelling in the study area, such trends point to possible future weakening of upwelling in the late summer-early autumn, consistent with climatic SST increase observed (Baptista et al., 2018). As some HAB dinoflagellates (Dinophysis, Lingulodinium) were observed in this study to be favored by weaker upwelling and warmer water in late summer and autumn, such climatic trends may increase the likelihood of their development. At the same time, upwelling conditions along the West coast of Iberia, including the study area, demonstrated a tendency to increase during colder half of the year, i.e., in period when water column is supposedly more mixed due to wind conditions. This may possibly promote more intensive phytoplankton blooms in spring related with upwelling events outsides of the main upwelling season. Such blooms are initially dominated by diatoms and flagellates in the study area (Loureiro et al., 2005; Goela et al., 2014), but at the later stages of succession dinoflagellates may be promoted, including HAB species e.g., as observed in spring 2016, when diatom bloom was followed by the appearance of HAB dinoflagellates (Danchenko et al., 2019).

Pronounced upwelling events in spring were observed repeatedly in this study, sometimes as early as February (e.g., in 1995, 1999 and 2000). Consecutively, the species of dinoflagellates that occurred after such short upwelling pulse has diminished (e.g., Dinophysis spp.) often resembled the community characteristic of the autumn succession in the end of upwelling season. Therefore, upwelling events during cold part of the year showed the potential to favor HAB species, which usually appear later during summer, and such prolonged presence of potential HAB organisms may pose a problem for aquaculture, as monitoring programs may require more frequent and longer closures of bivalve mollusk collection.