Yeosu Bay is located in the south-central region of the Korean Peninsula (Figure 1) and is connected to Gwangyang Bay through the Yeosu Channel. Gwangyang Bay is generally shallow, with a water depth of 2.4–8.0 m compared with the water depth of 30 m in Yeosu Bay (Kim et al., 2014; Choo, 2020). The Seomjin River and Gwangyang Bay are major introducers of nutrients and SPM to the study area (Kwon et al., 2001). However, Kang et al. (2019; 2020a; 2020b) reported that the main component of SPM is a phytoplankton-based autochthonic source, although SPM is gradually increasing, owing to artificial activities such as dredging and reclamation in Gwangyang Bay. The southern part of Yeosu Bay is connected to the South Sea of Korea and is seasonally affected by the Tsushima Warm Current. The tidal cycle is semidiurnal, with a maximum tidal range of 3.40 m during spring tide and 1.10 m during neap tide (Shaha and Cho, 2009). In summer, 300–400 m³ s^-1 of the Seomjin River water is continuously discharged into Gwangyang Bay and the Yeosu Channel, with 81.5% outflowing to the outside of Yeosu Bay through the Yeosu Channel (Kim et al., 2014). In this study, 15 stations were chosen in Gwangyang Bay in the Seomjin River Estuary (Stations 1–4) and Yeosu Bay (Stations 5–15).

Data were collected in June and August 2018, before and after rainfall. We considered the salinity gradient from the downstream entrance of the Seomjin River to the southern end of Geumo Island using the shipping vessel R/V Cheong-Gyeong-Ho from the Chonnam National University (Figure 1). Sea surface temperature (SST) and salinity data were collected using a multiparameter water quality measurement device (Water quality multi-parameter Pro DSS; YSI, Ohio, USA) in June, and a conductivity, temperature, and depth profiler (CTD) (SBE 19plus V2; Sea-bird Electronic, Washington, USA) in August. Although salinity measurements are unitless as indicated by recent studies (Williams et al., 2015; Gustafson et al., 2022; Wang et al., 2022), we used “psu” for distinguishing from classical unit (‰). Chlorophyll-a (Chl-a) concentration was obtained at each station by collecting 500 ml of surface seawater using a plastic bucket and filtering it through Whatman glass fiber filters {GF/F 47 mm; Whatman, Birmingham, UK). The filtered glass fibers were frozen until measurement. After adding 90% acetone to the frozen glass fiber filter, the pigment was extracted for 24 h.

Chl-a concentrations were measured using a spectrophotometer (Mega-800; Scinco Co., Seoul, Korea), according to the method described by Parsons et al. (1984). For SPM measurements, a glass fiber filter {GF/F 47 mm; Whatman, Birmingham, UK) was washed with distilled water in the laboratory and predried at 60°C for 24 h in a drying oven. The dried filter paper was weighed using an analytical balance (PR124/E; OHAUS Corp., New Jersey, USA) prior to sampling. In situ, 500 ml of surface seawater was collected in a plastic bucket, filtered through GF/F, frozen at –20°C, and transported to the laboratory. The dry weight of SPM was measured by drying the transported filter at 60°C for 24 h in a drying oven and weighing it (Korea’s Ministry of Environment, 2018). Owing to the possible effects of rainfall in the upper (Gurye) and lower (Hadong) reaches of the Seomjin River on the zooplankton community in Yeosu Bay, relevant data from these two locations were extracted from the Water Resources Management Information System (Figures 1B, C).

Owing to the large variation in zooplankton abundance, we performed sampling when the sea level was the highest during the spring tide period. Zooplankton were collected vertically using a conical net (net mouse diameter: 45 cm; mesh size: 200 µm). To calculate the filtered seawater volume, a flowmeter (Model 488115; Hydro-Bios Co., Altenholz, Germany) was attached to the entrance of the net. The collected zooplankton were immediately fixed in the field with a sodium tetraborate–buffered formaldehyde solution at a final concentration of 5%. Zooplankton samples were subsampled using a Folsom-type splitter of 1/8–1/128. Species identification of the zooplankton samples was performed where possible using a stereo microscope (SMZ645; Nikon, Tokyo, Japan), and the microstructure of appendages was observed using a high-magnification optical microscope (Eclipse E200; Nikon, Tokyo, Japan). Zooplankton samples were converted into individuals per cubic meter (inds. m^-3). Species were identified according to Chihara and Murano (1997); Soh (2010); Soh et al. (2013), and Soh and Moon (2014), and taxonomic systematics followed the WoRMS Editorial Board (2022).

Data converted to the log (x+1) index were used to normalize the total zooplankton abundance. For environmental factors (water temperature, salinity, Chl-a concentration, and SPM), Z-score standardization was performed to remove the units. Hierarchical cluster analysis was performed according to the Ward link method based on the Bray–Curtis dissimilarity index to measure similarities in zooplankton communities between stations (Legendre and Legendre, 1998). In addition, cluster analysis was performed using the nMDS (non-metric multidimensional scaling) ordination method (Clarke, 1993).

Using the cluster analysis results, an indicator value analysis was performed to identify the indicator species that affected each classified group (Dufrêne and Legendre, 1997). Subsequently, canonical correspondence analysis (CCA) and correlation analysis were performed by converting the indicator species abundance data to log (x+1) + 1 because of the exclusive species abundance by the assemblage group.

Bio-ENV analysis was performed to detect significant correlations between indicator species abundance and environmental variables (de Carvalho et al., 2015). A t-test was performed to identify the differences in environmental factors between the non-rainy period (June) and immediately after rainfall (August), and then, the differences in total zooplankton abundance were confirmed using the Mann–Whitney U test. In addition, the feeding habits of the zooplankton observed in Yeosu Bay in June and August were obtained from previous studies (Table 1). A chi-square test was performed to confirm the differences in the zooplankton feeding habit composition according to period. All data analyses were performed using the R program (version 4.1.0, R Core Team, 2020), whereas another graphing was performed using the “ggplot2” package; indicator species analysis using the “indicspecies” package; and CCA, correlation analysis, and Bio-ENV analysis using the “vegan” package.

The SST ranged from 22.5°C to 26.0°C before the rainfall and from 26.3°C to 28.0°C after rainfall (Table 2). Both before and after rainfall, the mean SST was higher in Gwangyang Bay than in Yeosu Bay. In Yeosu Bay, the SST was relatively low in the southwest regions (Figures 2A, B). Surface salinity ranged from 28.3 to 30.7 psu before the rainfall and from 20.8 to 32.8 psu after rainfall (Table 2). There were few differences in the salinity gradient between Gwangyang Bay and Yeosu Bay before the rainfall, but after rainfall, the salinity gradient extended to Yeosu Bay (Figures 2C, D).

SPM ranged from 0.4 to 18.1 mg L^-1 before the rainfall and from 1.6 to 68.8 mg L^-1 after rainfall (Table 2). There were no significant spatial differences, except in Station 2, before the rainfall. After rainfall, the SPM gradually decreased with increasing distance from the entrance of the Seomjin River estuary, and its tendency was similar to the distribution of the surface salinity gradient (Figures 2E, F). The Chl-a concentration ranged from 1.7 to 7.4 μg L^-1 before the rainfall and from 0.5 to 8.5 μg L^-1 after rainfall (Table 2). The Chl-a concentration was higher in the northeast of Yeosu Bay and lower in the center of Yeosu Bay before the rainfall, whereas it was higher in the central part of Yeosu Bay and was uniformly distributed after rainfall (Figures 2G, H).

The t-test analysis revealed that the average salinity before rainfall was significantly higher than that after rainfall (p<0.001). Water temperatures and the SPM were higher after rainfall than before rainfall (p<0.05 and p<0.001, respectively). The Chl-a concentration was not significantly different before and after rainfall (p>0.05). The t-test analysis revealed that the average salinity before rainfall was significantly higher than that after rainfall (p<0.001). SST and SPM were higher after rainfall than before rainfall (p<0.05 and p<0.001, respectively).

Four taxonomic groups (Appendicularians, Branchiopods, Copepods, Chaetognaths) were present before rainfall, and among these, only branchiopods were absent after rainfall (Figure 3). Appendicularian abundance ranged from 2.6% to 17.1%, except for Station 7 (no occurrence), before the rainfall, and in some transition zones (Stations 6–9) ranged from 0.4% to 2.8% after rainfall (Figure 3B). The proportion of appendicularians was higher in Yeosu Bay than in Gwangyang Bay before rainfall (Figure 3B). Branchiopod abundance ranged from 0.3% to 15.9% before rainfall and did not occur after rainfall (Figure 3). Branchiopods occupy a higher proportion in Yeosu Bay than in Gwangyang Bay. Copepod abundance ranged from 58.7% to 96.6% before rainfall and from 79.7% to 98.0 % after rainfall (Figure 3). Copepods were the most dominant among the zooplankton taxa, accounting for more than 58% of the total zooplankton abundance both before and after rainfall. Chaetognath abundance ranged from 0.3% to 29.8% before the rainfall and from 2.0% to 20.3% after rainfall (Figure 3). The proportion of chaetognaths before rainfall was higher in Yeosu Bay than in Gwangyang Bay, whereas it tended to be higher in Gwangyang Bay than in Yeosu Bay (Figure 3B).

The zooplankton abundance ranged from 416 to 19,975 inds. m^-3 before rainfall, and 617–3,582 inds. m^-3 after rainfall (Figure 3A). A significant decrease was observed in the total zooplankton abundance after rainfall (mean: 405±833 inds. m^-3) than that before rainfall (mean: 141±331 inds. m^-3) (Figure 3A, p<0.001, Mann–Whitney U test). Copepod abundance significantly contributed to the decrease in the total zooplankton abundance.

Acartia omorii, Paracalanus parvus s. l., Corycaeus spp., and copepodites were predominant before rainfall, and Bestiolina coreana, Centropages dorsispinatus, P. parvus s. l., and copepodites after rainfall (Figure 4). Copepodites were the most dominant taxa both before and after rainfall. The proportions of A. omorii and P. parvus s. l. tended to decrease from Gwangyang Bay to Yeosu Bay before rainfall, whereas the proportions of copepodites and P. parvus s. l. tended to increase from Gwangyang Bay to Yeosu Bay after rainfall (Figure 4). Zooplankton were distinguished by their two feeding habits (particle feeders, PFs; carnivores) before and after rainfall (Figure 4B). Before the rainfall, PFs ranged from 59.4% to 87.0%, and the carnivore proportion ranged from 13.0% to 40.6% (Figure 4B). After rainfall, the proportion of PFs increased by 70.3–94.9% and that of carnivores decreased by 5.1–29.7% (Figure 4B). PF accounted for more than 50% of the total zooplankton abundance both before and after rainfall. In addition, a significant increase in the PF proportion (Table 3, p<0.001) and a significant decrease in the carnivore proportion before and after rainfall were confirmed (Table 3, p<0.001).

Before rainfall, the carnivore abundance was positively correlated with SST, salinity, SPM, and Chl-a concentration. PF abundance also showed a positive correlation with all environmental variables, and the relationship with SPM was highly significant (Table 4, p<0.001). After rainfall, the carnivore abundance was positively correlated with SST, SPM, and Chl-a concentration but negatively correlated with salinity. However, water temperature and Chl-a concentrations were the only significant factors (Table 4; p<0.01 and p<0.05, respectively). Except for the relationship with salinity, which was not statistically significant (Table 4, p>0.05), the other factors exhibited a positive correlation (p<0.05). This correlation result suggests that salinity caused a reduction in the carnivore abundance, and prey factors (SPM and Chl-a concentration) were affected by an increase in PF abundance.

Based on Bray–Curtis dissimilarity–based cluster analysis, the study area was divided into three groups by zooplankton assemblages before rainfall (A, B, and C) and two groups after rainfall (A and B). IndVal analysis was performed only in Groups A and C before rainfall and Groups A and B after rainfall (Figure 5; Table 5). Although Group B included several stations before rainfall event, there were no species with an indicator power index greater than 25. Neritic species (A. erythraea, A. hongi, Evadne nordmanni, Labidocera rotunda, and Tortanus forcipatus) were extracted from both Groups A and C before rainfall, and they were statistically insignificant (Table 5). After rainfall, Group A comprised neritic and brackish species (A. erythraea, A. ohtuskai, A. sinjiensis, B. coreana, Parvocalanus crassirostris, and Pseudodiaptomus marinus), whereas Group B mainly comprised offshore warm-current species (A. pacifica, Canthocalanus pauper, C. furcatus, and Temora discaudata) in Yeosu Bay (Table 5).

The accumulated contribution of the two axes dividing Groups A and C was 83.86% before rainfall. The factors that had a significant correlation with neritic indicator species (E. nordmanni, A. hongi, A. erythraea, and T. forcipatus) before rainfall were surface water temperature and surface salinity (p<0.05) (Figures 6A, C). However, they did not significantly correlate with other environmental factors (Figures 6A, C). Indicator species showed a positive correlation with water temperature and a negative correlation with salinity (p<0.05) (Figure 6C). After rainfall, the accumulated contribution of the two axes, dividing groups A and B, was 86.62% (Figure 6B). Indicator species were consisted of A. erythraea, B. coreana, A. ohtsukai, Parvocalanus crassirostris, Pseudocalanus marinus, and A. sinjiensis for Group A and Calanus sinicus, Canthocalanus pauper, Temora discaudata, C. furcatus, and A. pacifica for Group B. The indicator species of Group A showed a positive correlation with SST and SPM and a negative correlation with salinity (p<0.05) (Figure 6D). The indicator species of Group B showed a positive correlation with salinity and a negative correlation with SST, Chl-a concentration, and SPM (p<0.05) (Figure 6D). Moreover, after rainfall, there was a high correlation between the indicator species and the environmental factors (Figure 6D, p<0.05).

Bio-ENV was used to analyze the combination of environmental factors that exerted the greatest effects on indicator species abundance, and the results indicated the highest correlation between surface water temperature and surface salinity before rainfall (correlation coefficient: 0.7576, Table 6). However, the highest correlation was observed for a single combination of surface salinity after rainfall (correlation coefficient: 0.7521, Table 6).