Data were collected during two “Acoustics along the Brazilian coast” surveys (ABRACOS 1 and ABRACOS 2; Bertrand, 2015, 2017) performed along the northeast Brazilian continental shelf and slope between 5 and 9∘ S and around oceanic seamounts and islands from FNR, including the Fernando de Noronha Archipelago itself and the Rocas Atoll (Fig. 1). The surveys were carried out on board the French oceanographic vessel R/V Antea in austral spring (September–October 2015) and autumn (April–May 2017). Planktonic cnidarians were sorted from zooplankton samples collected at 34 and 45 stations in spring and autumn, respectively (Fig. 1). Samples were collected through oblique hauls, with a bongo net with a 300 µm mesh size and 0.6 m mouth opening. The water column was sampled from near the bottom to the surface over the continental shelf and from 200 m to the surface in the offshore. The net was towed at approximately 2 knots, at various times of day and night. The net was fitted with a calibrated mechanical flowmeter (Hydro-Bios) to estimate the volume filtered during each haul. Samples were fixed with 4 % formaldehyde buffered with sodium tetraborate (0.5 g L-1).

In the laboratory, whole zooplankton samples were analysed under a stereomicroscope, and cnidarian specimens were identified (mainly according to Pugh, 1999; Bouillon, 1999) and counted. Abundances were standardised with regard to the number of individuals (ind.) per 100 m-3 for medusae and number of colonies per 100 m-3 for siphonophores. For calycophorans, the number of anterior nectophores was used for estimating the polygastric stage abundance and eudoxid bracts for the eudoxid stage abundance (e.g. Hosia and Båmstedt, 2007; Hosia et al., 2008a). For physonects and Hippopodidae, the number of colonies was roughly estimated by dividing the number of nectophores by 10 (Pugh, 1984).

Vertical profiles of temperature (∘C), practical salinity and fluorescence were obtained with a CTD-O2 profiler Seabird SBE911+. Conductivity, temperature and pressure accuracies were estimated at 0.0003 S m-1, 0.001 ∘C and 0.7 dbar, respectively. In addition to CTD-O2, continuous along-track sea surface Practical Salinity data were acquired with a thermosalinograph SBE21. Along-track current profiles were recorded with an “Ocean Surveyor” ship-mounted acoustic Doppler current profiler (SADCP) operating at a frequency of 75 kHz with a depth range of 15–700 m. SADCP data were processed and edited using the Common Ocean Data Access System (CODAS) software package developed at the University of Hawaii (https://currents.soest.hawaii.edu/home/index.html, last access: 15 January 2018). The relative velocities were rotated from the transducer to the Earth reference frame using the ship gyrocompass. The global positioning system (GPS) was used to retrieve the absolute current velocities. The orientation of the transducer relative to the gyroscopic compass and an amplitude correction factor for the SADCP were determined by standard calibration procedures. Finally, velocity profiles were averaged hourly, providing profiles in the 19–600 m range. SADCP data over the shelf (bathymetry shallower than 70 m) were often contaminated by spurious reflections from the bottom, so the data coverage was only partial in these areas. The current measurements were resampled to 0.1∘ spatial resolution and depth integrated (0–70 m depth). The analyses of the data obtained by SADCP are discussed in Dossa et al. (2021) and Costa da Silva et al. (2021).

For the data analysis, stations were grouped in three systems (Fig. 1): continental shelf, WBCS (offshore stations adjacent to coast, mainly over the slope) and the South Equatorial Current system (SECS; stations along the FNR). WBCS and SECS present a distinct thermohaline structure (Assunção et al., 2020). Stations A2_39 and A2_54 were located in the transition zone between WBCS and SECS and thus removed for the comparative analysis. To dampen effects of dominant species, abundance data were transformed by log (x+1) in all analyses. A permutational multivariate analysis of variance (PERMANOVA; Anderson et al., 2008) was used to test for diel differences in the community structure of planktonic cnidarians on both seasons. Since no significant differences were observed (pseudo F=0.93842 and 0.86812, p=0.451 and 0.509, for spring and autumn, respectively) the pooled set of day and night data was used.

Factorial analysis of variance (ANOVA) was performed to test for differences in hydromedusae and siphonophores richness and total abundance and the abundance of the most abundant species (abundance >1 %), according to the region (shelf, WBCS and SECS), the season and the interactions between these factors. When ANOVA was significant, a Tukey post hoc test was used to identify the levels (seasons and regions) that differed among each other.

Spatial patterns in planktonic cnidarian community abundance in autumn were identified by hierarchical cluster analysis using Bray–Curtis similarity matrix. The validity of the groups defined by the cluster analysis was tested through the SIMPROF test (5 % significance level). A similarity percentage (SIMPER) analysis was performed to identify representative species and their contribution to similarity within the groups defined by the cluster analysis. This procedure was previously performed by Tosetto et al. (2021) for the spring dataset.

For each system, constrained ordination analyses were performed to identify associations between the most abundant planktonic cnidarian species and the environmental variables. The following continuous explanatory variables were used: (i) surface layer (0–30 m) temperature, (ii) salinity, (iii) dissolved oxygen (except for shelf), (iv) fluorescence (as an indirect measure of biological productivity), (v) mixed-layer depth, (vi) bottom depth, (vii) the zonal component of acoustic Doppler current profiler (ADCP) data integrated over the first 70 m depth and (viii) the meridional component of ADCP data integrated over the first 70 m depth. Seasons (spring and autumn) were used as qualitative explanatory variables. Detrended canonical correspondence analysis (DCCA) revealed a small length of variable gradients (>3), indicating that a linear method was more appropriate to use on this occasion, and thus redundancy analysis (RDA) was selected (Lepš and Šmilauer, 2003).

Distribution maps were produced with QGIS 3.20 (QGIS Development Team, 2022). Factorial ANOVA was performed in Statistica 10 (StatSoft Inc., 2011). Cluster, SIMPROF, SIMPER and PERMANOVA analyses were performed in Primer v.6 + PERMANOVA (Clarke and Gorley, 2006). DCCA and RDA were performed in CANOCO 4.5 (Lepš and Šmilauer, 2003).

Whatever the season, surface (0–70 m) SADCP data showed the central branch of the South Equatorial Current (cSEC) flowing westward in the open ocean with its core around FNR (Fig. 2a, b). Meanwhile, the NBUC flows northward over the continental slope with intrusions over the continental shelf. The cores of both currents meet up in the northwest part of the study area (around 3.5∘ S, 35∘ W) forming the North Brazil Current (Bourles et al., 1999; Dossa et al., 2021; Stramma and England, 1999).

Although the NBUC was weaker south of 7.5∘ S in both seasons, it was even weaker in spring 2015 (Fig. 2a) due to the presence of an anticyclonic eddy. Thus, intrusions of oceanic waters over the shelf were greatly reduced in the region of the Pernambuco Plateau in spring (Fig. 2a; Tosetto et al., 2021; Dossa et al., 2021). Such a feature was not observed in autumn, and thus intrusions over the shelf were observed all along the coast (Fig. 2b). In deeper waters, we observed the core of the NBUC flowing along the entire extent of the slope in both seasons (Fig. 2c, d). However, it was stronger and deeper (upper limit at ∼105 m; Fig. 2c), in spring than in autumn, when the core was less intense, shallower (upper limit at ∼80 m) and flowing slightly further away from the coast (Fig. 2d). Such differences were reflected in surface currents, which were stronger over the slope in autumn (Fig. 2b) but similar in both seasons over the continental shelf. (Fig. 2a, b; Dossa et al., 2021). For a detailed description and discussion of circulation patterns and seasonal variability of the NBUC, see Dossa et al. (2021).

In both seasons, the WBCS (over the slope) and the SECS (FNR) presented a distinct thermohaline structure (Fig. 3b, c, e, f). The thermocline and halocline were sharp in the SECS (Fig. 3c, f), with the South Atlantic Central Water mass (SACW; <13 ∘C) reaching up to ∼150 m depth. They were less abrupt in WBCS (Fig. 3b, c) and the SACW was only observed below 250 m (Fig. 3 in Silva et al., 2021; Fig. 3 in Dossa et al., 2021). In both offshore areas, the upper limit of the mixed-layer depth (MLD) was shallower in autumn. However, the difference was much more pronounced in the SECS (90 and 46 m in spring and autumn, respectively; Fig. 3c, f) than in the WBCS (53 and 39 m in spring and autumn, respectively; Assunção et al., 2020).

Sea surface temperature was higher in autumn in offshore areas and over the continental shelf (Figs. 3, S1 in the Supplement). Due to the higher temperature, evaporation was intense and sea surface salinity was generally higher during autumn even with the greater rainfall during this season (Figs. 3, S1). Specifically, over the continental shelf, distinct patterns were observed in sea surface salinity from thermosalinograph data during each season (Fig. S2). In spring, as a result of the reduced continental discharge and higher evaporation over shallow areas, surface salinity was higher in the more inshore areas, with a reduction when approaching the shelf break and offshore. Differently, in autumn, although the overall salinity was higher, the highest values were along the outer shelf/shelf break, with lower salinity in inshore waters (up to ∼30 m isobath) due to the higher continental runoff (Fig. S2). For a detailed description and discussion of the seasonal variability of the thermohaline structure, see Assunção et al. (2020).

Over the continental shelf, although presenting high variability, fluorescence was higher in near-bottom water and in the surface layer in autumn (Fig. 3g), coinciding with a period of larger continental discharge and nutrient input. In both the WBCS and the SECS, the peak of fluorescence reached similar values within the thermocline in both seasons (Fig. 3h, i). However, in the surface layer, while in the WBCS fluorescence was similar in both seasons (note the small difference at the MLD), in the SECS it was higher in autumn when the MLD and the nutricline were shallower (Fig. 3h, i; Assunção et al., 2020; Silva et al., 2021; Farias et al., 2022).

In total, 93 taxa of planktonic cnidarians were sampled in the area, corresponding to 42 hydromedusae, 48 siphonophores, and 3 scyphomedusae. A similar number of taxa were present at each season: 73 in spring and 74 in autumn, with 18 and 19 exclusive species, respectively. In addition, many unidentified larval stages (cerinula, ephyrae and athorybia) were collected (Table S1 in the Supplement; Table 1 in Tosetto et al., 2021). In both seasons, the most frequent and abundant species were almost the same. Among hydromedusae, Aglaura hemistoma and L. tetraphylla occurred in more than 85 % of the samples. However, while A. hemistoma was the most abundant in spring, L. tetraphylla dominated in autumn. Among siphonophores, Abylopsis eschscholtzii, Abylopsis tetragona, B. bassensis, C. appendiculata, D. bojani, Eudoxoides mitra and Sulculeolaria chuni were present in more than 80 % of the samples in both seasons and Eudoxoides spiralis in spring. B. bassensis, C. appendiculata, D. bojani and E. mitra were also the most abundant siphonophores on both cruises, representing 46.5 % of the total abundance in spring and 36.9 % in autumn (Table S1, Table 1 in Tosetto et al., 2021).

Hydromedusae species richness was similar all over the study area, with no significant differences among areas or seasons and averaging 4.5 ± 2.3 and 4.1 ± 1.5 species per station in spring and autumn, respectively (Table 1, Fig. 4). However, while in spring the total hydromedusae abundance was similar in all areas, averaging 49.9 ± 66.2 ind. 100 m-3, in autumn it was significantly higher over the continental shelf, where it reached up to 1067.6 ind. 100 m-3 (383.4 ind. 100 m-3 on average; Table 1, Figs. 4, 5; Fig. 2 in Tosetto et al., 2021). No significant seasonal changes were observed in total hydromedusae abundance in the open ocean.

The two dominant hydromedusae, L. tetraphylla and A. hemistoma, were widespread over the area. L. tetraphylla presented higher abundance over the continental shelf in both seasons. However, in autumn it was 1 order of magnitude higher than in spring on average, reaching up to 770.3 ind. 100 m-3 (Table 1, Figs. 5, 6; Fig. 3 in Tosetto et al., 2021). In both seasons, A. hemistoma was more evenly distributed than L. tetraphylla, with high abundances occurring both over the continental shelf and in the open ocean. But, as occurred with L. tetraphylla, it was significantly higher over the continental shelf in autumn (Table 1, Figs. 5, 6; Fig. 3 in Tosetto et al., 2021). The third most abundant hydromedusa taxon, Aequorea spp., occurred in low abundances over the area in spring. However, in autumn its abundance was significantly higher in the SECS, reaching up to 38.3 ind. 100 m-3, and was nearly absent over the shelf and in the WBCS (Table 1, Figs. 5, 6).

Siphonophores richness was significantly higher in the two oceanic areas than over the continental shelf, especially in the WBCS where it was also significantly higher in spring (reaching 23 species, 19.5 ± 2.9 on average) than in autumn (13.3 ± 1.5 species on average). No differences among seasons were observed over the shelf and in the SECS (Table 1, Fig. 4). The total abundance of siphonophores was similar between seasons over the continental shelf (137.2 and 169.5 ind. 100 m-3 on average in spring and autumn, respectively). Meanwhile, a contrasting pattern was observed in the oceanic areas: in spring it was higher in the WBCS (not statically significant), averaging 153.4 ind. 100 m-3, while in autumn significantly higher abundances were observed in the SECS (245.2 ind. 100 m-3 on average; Table 2, Fig. 4; Fig. 2 in Tosetto et al., 2021).

Although most dominant siphonophores were widespread over the area, distinct seasonal and spatial patterns were observed. E. mitra was more abundant in the SECS in both seasons. However, the pattern was more pronounced in autumn when this species reached up to 109.5 ind. 100 m-3. A. eschscholtzii, which occurred in similar abundances in the three areas in spring, was significantly more abundant in the SECS in autumn, reaching up to 67.6 ind. 100 m-3. In the WBCS, average abundances of E. mitra and A. eschscholtzii were similar in spring and autumn (Table 1, Figs. 5, 6; Fig. 4 in Tosetto et al., 2021). Otherwise, over the continental shelf and in the WBCS, D. bojani and B. bassensis were more abundant in spring than in autumn, while in the SECS the opposite was observed, and these species were significantly more abundant in autumn (Table 1, Figs. 5, 6; Fig. 4 in Tosetto et al., 2021).

In spring, C. appendiculata was slightly more abundant over the continental shelf and in the WBCS than in the SECS. Although, in autumn its abundance was higher over the continental shelf and lower in the two oceanic areas, seasonal changes were not significant (Table 1, Figs. 5, 6; Fig. 4 in Tosetto et al., 2021). Otherwise, A. tetragona and E. spiralis were more abundant over the continental shelf and in the WBCS than in the SECS in spring. Conversely, in autumn they were less abundant in the shelf and WBCS than in autumn, and the abundances of these species were similar in the three systems (Table 1, Figs. 5, 6; Fig. 4 in Tosetto et al., 2021).

Distinctly from other dominant siphonophores, in both seasons M. kochii occurred almost exclusively over the continental shelf. However, while in spring the species was restricted to few stations in the south of the area, in autumn it occurred in higher abundances and was distributed over most of the continental shelf, except a few stations in the outer shelf (Table 1, Figs. 5, 6; Fig. 4 in Tosetto et al., 2021). No spatial and temporal patterns were identified in the abundance of S. chuni.

In spring, two main groups of stations were depicted in the cluster analysis. Group A was represented by four stations over the shelf in the southernmost part of the study area (Fig. 7b). It was mainly characterised by the large abundances of M. kochii, A. hemistoma and L. tetraphylla. It also differed from the other groups by the absence or low abundance of other dominant siphonophores. Group B was represented by the remaining neritic stations and all stations of the WBCS and the SECS where siphonophores such as D. bojani, B. bassensis, A. tetragona, C. appendiculata and E. mitra were more abundant. This group was further subdivided into several subgroups, which differed among each other in the abundance of dominant species. For a detailed description and discussion of the cnidarian community structure in the area in spring, see Tosetto et al. (2021).

Similarly, two main groups were depicted in autumn (Fig. 7c). Group X was composed by shallow stations (bottom depth <25 m) over the continental shelf. SIMPER analysis indicated an average similarity of 63.1 %. It was mainly characterised by the occurrences in large or medium abundance of L. tetraphylla, M. kochii, A. hemistoma, A. eschscholtzii, C. appendiculata and B. bassensis. One outlier was present in this branch (station autumn_3), and it differed from the main branch due to the absence of A. hemistoma, A. eschscholtzii, C. appendiculata and B. bassensis (Table 2).

Group Y was composed by the deeper stations over the continental shelf and all stations in the WBCS and the SECS. This group was subdivided into two large branches and several outliers. Subgroup Y1 was composed by most stations of WBCS over the slope. SIMPER analysis indicated an average similarity of 73.4 % within the group (Table 2). It was represented by the species C. appendiculata, D. bojani, E. mitra, B. bassensis, A. tetragona, A. hemistoma, L tetraphylla, A. eschscholtzii, Lensia meteori and E. spiralis, occurring in high or medium abundances. Subgroup Y2 (average similarity of 78.1 %) included most stations in SECS, represented by the species D. bojani, B. bassensis, E. mitra, A. eschscholtzii, A. hemistoma, C. appendiculata, A. tetragona, S. chuni, Cordagalma ordinatum and Aequorea spp (Table 2). Subgroup Y2 differed from Y1 due to the higher abundance of A. eschscholtzii, B. bassensis, Aequorea spp., D. bojani, A. hemistoma, E. mitra and C. ordinatum and the lower abundance of L. meteori and C. appendiculata.

The two first canonical axes of the RDA explained 45.9 %, 28 % and 51.4 % of species variance for the shelf, WBCS and SECS, respectively (Table 3). A monte Carlo test showed that the first and all canonical axes together were significant (p<0.05) for the three analyses. Over the continental shelf, the first axis was related to the cross-shelf spatial structure, being positively related to fluorescence and negatively related to bottom depth. The second axis was related to seasonal variability being positively related to autumn, with stronger northward surface currents and higher temperature, salinity and fluorescence, and negatively related to spring (Fig. 8a, Table 3). In this system, L. tetraphylla and M. kochii were positively related with axis 1, such a relation reflecting their higher abundances in shallower and more productive coastal waters during both seasons. Other species were negatively related with axis 1, reflecting their higher abundances in the areas over the shelf with the influence of oceanic waters. Among these species, A. eschscholtzii and D. dispar were also positively related to axis 2, reflecting their higher abundance during autumn, while B. bassensis, A. tetragona and E. spiralis were negatively related to axis 2, reflecting their higher abundance in spring (Fig. 8a).

In the WBCS, the first axis was mainly related to the seasonal variability, being positively related to spring, with a deeper mixed layer and weaker westward currents or the presence of eastward surface currents (see Tosetto et al., 2021), and negatively related to autumn, with higher temperature and fluorescence. The second axis was positively related to salinity and negatively related to bottom depth and stronger northward surface currents (Fig. 8b, Table 3). In this system, C. appendiculata, B. bassensis, A. tetragona, E. Spiralis and L. meteori were positively related to axis 1, reflecting their higher abundance during spring. E. mitra, D. bojani, D. dispar and A. eschscholtzii were negatively related to axis 1, reflecting their higher abundance in autumn, when shallower mixed layers and higher fluorescence and temperature were observed. A. hemistoma, L. tetraphylla and S. chuni were positively related to axis 2 and shallower waters over the slope (Fig. 8b).

In the SECS, the first axis was mainly related to the seasonal variability being positively related to autumn, with higher temperature and fluorescence, and negatively related to spring, with a deeper mixed layer and weaker westward surface currents. The second axis was positively related to stronger northward surface currents (Fig. 8c, Table 3). In this system, E. mitra, A. hemistoma, A. tetragona, S. chuni, B. bassensis, D. bojani and C. ordinatum. A. eschscholtzii, D. dispar and Aequorea spp. were positively related to axis 1, reflecting their higher abundance in autumn, when shallower mixed layers and higher fluorescence and temperature were observed. Meanwhile, L.tetraphylla, E. spiralis, C. appendiculata and L. meteori were negatively related to axis 1, reflecting their higher abundance in spring (Fig. 8c).