Field sampling was conducted at 13 stations (S1–S13) along the southern KCWs in March and April 2021 ( Figure 1A ). In situ measurements of water quality parameters, including temperature, salinity, pH, and dissolved oxygen (DO), were performed at the surface water using a device from YSI EXO2 Sonde probes (Yellow Springs, OH, USA). Surface water samples were collected using a bucket at all sampling stations. Subsamples (500 mL) that were used for analysis of inorganic dissolved nutrients and chlorophyll a (Chl. a) were immediately filtered through GF/F filters (diameter: 47 mm, pore size: 0.7 μm; Whatman, UK) in the field. The filtrates were then transferred to acid-cleaned 15-mL conical tubes (SPL Life Sciences, Korea), and HgCl₂ (final concentration 0.1%) was added to prevent changes from biological reactions (Kirkwood, 1992). The filters and filtrates were stored at −20°C in the dark prior to laboratory analysis. For analysis of phytoplankton composition, 500 mL of surface water was collected in polyethylene (PE) bottles and immediately fixed with 3% Lugol’s solution. Data on precipitation (Geoje Island) and solar radiation (Namhae) for the study area during 2021 were from the Korea Meteorological Administration (KMA, www.weather.go.kr).

The mesocosm experiment was from 24 March to 8 April 2021 in the coastal waters of Geoje Island, South Korea (34°59′37″N, 128°40′27″E). Each mesocosm was cylindrical (diameter: 1.0 m, height: 2.5 m), made of transparent PE, and contained 1000 L of natural seawater that was continuously collected from below 2 m using polyvinyl chloride (PVC) and PE hoses and a pumping apparatus ( Figure 1B ). To prevent accidental entry of surrounding seawater due to wave action, the upper 1 m of each mesocosm extended above the water surface. All mesocosms were securely tied to a steel structure using ropes. The control mesocosms had natural seawater alone, and each treatment group had N:P:Si additions in a molar ratio of 16:1:16, the Redfield ratio (Redfield et al., 1963); the low nutrient (LN) group had 32 μM nitrate, 2 μM phosphate, and 32 μM silicate And the high nutrient (HN) group had 96 μM nitrate, 6 μM phosphate, and 96 μM silicate. Sodium nitrate, sodium metasilicate nonahydrate, and sodium phosphate from JUNSEI (Japan) were the sources of nitrate, silicate, and phosphate. There were six mesocosms, with two mesocosms per group. Due to the large size and handling limitations (non-uniformity within repetitions) of mesocosm experiments, numerous researchers have conducted experiments utilizing gradient conditions rather than three repetitions, and the validity of their results has been demonstrated in many publications (Schä et al., 2001; Sandaa et al., 2009; Zöllner et al., 2009; Baek et al., 2013; Park et al., 2014; Lee et al., 2023). Considering this, the duplicated gradient experiments were performed in this study. The water column within each mesocosm was mixed once a day and before sample collection using a 3.5-m long vertical agitator made of PVC.

The physicochemical conditions during the experiment (temperature, salinity, DO, and pH) were measured at the water surface in each mesocosm using a YSI EXO2 Sonde probes (Yellow Springs, OH, USA). Surface water samples were also manually collected using a non-toxic PE beaker after whole water mixing in mesocosm on day 0, 2, 4, 6, 8, 10 and 15. Similar to the field sampling, mesocosm samples were collected for measurement of inorganic nutrients and Chl. a. For identification and enumeration of phytoplankton, water samples were fixed with 3% Lugol’s solution and stored at room temperature until analysis.

For laboratory analysis of field and mesocosm samples, the concentration of Chl. a (µg L^-1) was determined using a Turner 10-AU Field Fluorometer (Turner BioSystems, Sunnyvale, CA, USA). Prior to analysis, the filtered material was extracted with 90% acetone and kept in the dark for 24 h. The concentrations of dissolved inorganic nutrients (ammonium, nitrate + nitrite, phosphate, and silicate) were measured using a flow injection auto-analyzer (Quattro 39; Seal Analytical, Fareham, Hampshire, United Kingdom), as described by Parsons et al. (1984), which was calibrated using Reference Materials for Nutrients in Seawater (RMNS; KANSO Technos Co., Ltd., Japan). To count and identify phytoplankton, each Lugol’s-fixed sample (500 mL) was concentrated to approximately 50 mL by decanting the supernatant and storage at room temperature. These sub-samples were allowed to settle for 10 min and then a sample of 100 to 400 µL was transferred into a Sedgewick-Rafter Chamber. Identification and counting of phytoplankton cells were performed using light microscopy (200 or 400×). Morphologically distinct species were identified to the species and genus levels (Wetzel and Likens, 2000; Tomas, 1997).

In addition to cell density, the biovolume of marine phytoplankton cells is important to the study of phytoplankton ecology (Sun and Liu, 2003). The dominant species (Chaetoceros spp., Cylindrotheca closterium, Eucampia zodiacus, Pseudo-nitzschia spp., and Cryptomonas spp.) cell density data in this study were converted to biovolume (μm³ L^-1) using cell dimensions applied to Olenina (2006) as cell density × cell volume of each species. The specific growth rate of dominant diatom in mesocosm experiment was calculated as Equation 1:

where X_t is the cell concentration at time t, X₀ is the initial cell concentration, and △t is the incubation time in days (Wood et al., 2005). The t-test was used to determine the significance of differences in abiotic and biotic factors between field stations during March and April, and a p-value below 0.05 was considered significant. Statistical analyses were performed using SPSS version 17.0 (SPSS Inc., Chicago, IL, USA).

For analysis of phytoplankton communities (group averages) in the field, non-metric multidimensional-scaling (MDS) ordination using the Bray-Curtis similarity index was performed to analyze species composition data with PRIMER 5 software. The relationships of phytoplankton groups with environmental factors in the southern KCWs and selection of variables that best described the distribution of groups were analyzed using redundancy analysis (RDA), a linear method of direct ordination (Ter Braak, 1994; Ter Braak and Smilauer, 1998). Specifically, RDA was used to determine the correlation of planktonic algae at the genus level with environmental factors in all mesocosms using CANOCO for Windows version 4.5, with log-transformation of phytoplankton numbers and environmental variables prior to analysis.

To assess the associations of dominant phytoplankton species with environmental factors in the mesocosms and to identify variables that had the strongest effect, principal component analysis (PCA), a linear method of direct ordination (Wold et al., 1987), was performed using R version 4.2.1.

Our analysis of data collected at 13 field stations demonstrated that the surface water temperature ranged from 10.04 to 12.25°C in March and from 13.16 to 17.04°C in April ( Figure 3A ). During both months, the surface salinity remained above 32 ( Figure 3B ) and the DO concentration ranged from 7.83 to 11.61 mg L^-1 ( Figure 3C ). The pH varied from 7.86 to 8.48 ( Figure 3D ). Among these environmental factors in March and April, only DO concentration had a significant difference (p< 0.05; Figures 3E−H ).

There were significant spatial variations in dissolved inorganic nutrients during March and April ( Figure 4 ). The nitrate+nitrite concentration varied from 1.62 to 10.98 μM in March and from 1.58 to 56.01 μM in April, and the highest concentration was at S6. The ammonium concentration ranged from 0.81 to 6.14 μM in March and from 0.55 to 6.42 μM in April, and the highest concentration was also at S6. The phosphate concentration had minimal fluctuations, and only ranged from 0.37 to 0.58 μM during both months. The silicate concentration remained relatively high, and ranged from 1.91 to 14.25 μM in March and from 3.74 to 44.23 μM in April. The concentrations of nitrate+nitrite and silicate were significantly higher in April compared to March (p< 0.05; Figures 4E−H ).

In March, the concentration of Chl. a ranged from 0.26 to 3.23 μg L^-1 (average: 0.89 ± 0.89 μg L^-1) and the average total phytoplankton abundance was 1.25 × 10⁵ ± 1.11 cells L^-1, with the highest levels of both parameters at S1 and S13 ( Figure 5A ). In April, the average Chl. a concentration was 1.20 ± 0.41 μg L^-1, the total phytoplankton abundance was 1.12 to 9.84 × 10⁵ cells L^-1, and both of these parameters were higher than during March ( Figure 5B ).

The dominant species at most stations during March were Eucampia zodiacus and Cryptomonas spp. ( Figure 5C ). In particular, Cryptomonas spp. accounted for nearly 50% of all phytoplankton at S6 and S7, which were near the river. During April, at S6 and S7, Cryptomonas spp. was dominant and accounted for more than 90% of all phytoplankton ( Figure 5D ); at S9, S12, and S13, Skeletonema spp. was dominant; and at S2, Chaetoceros spp. accounted for approximately 52% of all phytoplankton.

The changes in water temperature and salinity were similar in the three different mesocosm groups ( Figures 6A, B ). In particular, the water temperature increased in all three groups throughout the experimental period, and ranged from 11.81 ± 0.01°C to 14.11 ± 0.05°C. The salinity slightly decreased to 31.78 ± 0.01 from 32.84 ± 0.04, and was lowest following rainfall on day 10. The DO concentration was lowest at the start of the experiment (9.13 ± 0.04 mg L^-1) and increased in all three groups until day 3 ( Figure 6C ); after day 3, the DO level decreased in the control mesocosm; on day 8, the DO concentration was 13.26 ± 0.23 mg L^-1 in the LN group and 18.92 ± 0.54 mg L^-1 in the HN group. The DO and Chl. a dynamics were similar in the HN group (R^{2 =} 0.80, p< 0.01), and each had a maximum on day 8. The pH ranged from 8.75 ± 0.01 to 9.11 ± 0.02, and the highest value (9.11) was on day 8 ( Figure 6D ), at which time there was a significant depletion of nutrients in the HN group.

In general, the nutrient concentrations gradually decreased over the 2-week experiment, and were mostly depleted by day 8 ( Figures 6E−H ). In the control group, the nitrate+nitrite concentration was 1.91 ± 0.08 μM initially, 2.94 ± 0.48 μM at the midpoint, and 1.04 ± 0.24 μM at the end ( Figure 6E ). The ammonium concentration remained relatively low in all three groups, and ranged from 0.13 to 1.33 μM ( Figure 6F ). In the control group, the silicate level varied from 1.58 to 2.91 μM ( Figure 6G ); in the LN group, the silicate level was 27.27 ± 2.01 μM initially, 0.50 ± 0.01 μM on day 8, and 3.40 ± 1.01 µM on day 15; in the HN group, the initially high silicate level decreased to 0.61 ± 0.03 μM on day 8 and remained low until day 15. In the control group, the phosphate concentration varied from 0.01 ± 0.01 to 0.12 ± 0.01 μM ( Figure 6H ); in the LN and HN groups, the phosphate level declined to undetectable levels on day 8.

We performed cluster analysis and non-metric MDS to analyze phytoplankton composition in southern KCWs from field data ( Figures 9A–D ). The results showed that the phytoplankton community structure during each month separated into three distinct groups (both p< 0.01). In March, the stations in group 1 (S1, S2, S3, S4, S10, S11, S12) were characterized by E. zodiacus and the stations in group 2 (S6, S9, S13) were characterized Cryptomonas spp. Notably, the stations within these groups were geographically distant from each other, indicating significant spatial variations in the composition of phytoplankton communities. In April, the stations in Group 1 (S5, S6, S7) were characterized by a dominance of Cryptomonas spp. and Skeletonema spp., and these three stations were close to the river, indicating that river effluent likely affected phytoplankton community composition.

We also performed RDA of the field data to identify the relationships of phytoplankton with environmental variables ( Figures 9E ). The results demonstrated that multiple variables (temperature, ammonium, nitrate+nitrite, salinity, silicate, phosphate, DO, pH, and Chl. a) explained a significant amount of the variability in the phytoplankton communities. The first two redundancy axes accounted for 52.5% of the variance in the species-environmental variable biplot (axis 1: 49.4%; axis 2: 22.5%). Axis 1 had a strong positive correlation with pH and a negative correlation with nitrate+nitrite and silicate. Axis 2 had strong negative correlations with the concentrations of DO and Chl. a, and a positive correlation with ammonia. In addition, Cryptomonas spp. had a negative association with salinity and E. zodiacus.

In mesocosm data, we used PCA to analyze phytoplankton composition ( Figure 10 ). The species-environment variable biplot demonstrated that the first two redundancy axes explained 67.6% of the total variance (PC1: 39.8%, PC2: 27.8%). PC1 had positive loading for Chl. a, Chaetoceros spp., and E. zodiacus, and negative loading for NH⁴⁺. PC2 had positive loading for the nutrient variables, and C. closterium had a negative correlation with the nutrient variables.

It is widely recognized that nutrient levels have major effects on the growth of phytoplankton in coastal waters (Wang et al., 2010; Baek et al., 2019; Baek et al., 2020; Youn et al., 2022). Nutrients are primarily introduced into these coastal waters by mixing, upwelling events, and river runoff, and these events are especially common during the rainy seasons in temperate regions (Baek et al., 2015; Baek et al., 2018; Baek et al., 2020). Our field measurements indicated the mean concentrations of dissolved inorganic nitrogen (DIN) and dissolved silicate (DSi) were significantly higher during April than March (DIN: 6.74 ± 4.02 vs. 15.18 ± 16.12 μM, t-test: t = minus;1.83, p< 0.05; DSi: 5.93 ± 3.97 vs. 11.95 ± 11.15 μM, t-test: t = −2.06, p< 0.05). There was also a significantly greater abundance of phytoplankton during April than March (t-test: t = −3.12, p< 0.05). The cumulative rainfall during the 7-day period prior to the sampling in April (79.6 mm) was approximately twice as high as in March (38.8 mm). Therefore, the greater rainfall during April led to increased discharge of nutrients from nearby rivers, and this was the most likely cause of the increased abundance of phytoplankton. The findings from our field survey, together with the results of our mesocosm experiments (described below), highlight the important effects of nutrient inputs from terrestrial runoff on phytoplankton populations in nearby coastal waters, even during the early spring. Our findings are also consistent with previous studies by Yoon et al. (2023) and Baek et al. (2015; 2020), which also demonstrated trends of nutrient-driven effects on phytoplankton population dynamics in mesocosm and field surveys during the summer.

In March, the dominant species following cell density were the diatom E. zodiacus and the cryptophyte Cryptomonas spp. Based on geographical data and MDS analysis, we divided the phytoplankton assemblage into three groups, and there were two main groups. Group I, which is characterized by E. zodiacus, and Group II, which is characterized by Cryptomonas spp. These two groups were in different regions of the southern KCWs, and had significant differences in silicate (t-test: t = 1.23, p< 0.05) and salinity (t-test: t = −2.19, p< 0.05). Group II was characterized by high silicate and low salinity, and cryptophytes, including Cryptomonas spp., are more tolerant of changes in salinity than diatoms and dinoflagellates (Thomas and Gibson, 1990; Yoon et al., 2022). Moreover, Cryptomonas spp. is dominant in many estuarine areas (Lee et al., 2020; Yoon et al., 2022). Therefore, the presence of cryptophytes in Group II was likely attributable to the low salinity due to terrestrial runoff. The RDA analysis supported a negative correlation between Cryptomonas spp. and salinity. The diatom E. zodiacus can utilize nitrogen efficiently at low temperature conditions (around 9°C) and, the blooms of E. zodiacus have occurred almost every year during the low temperature period (7.5 to 11.0°C) from winter through early spring (Nishikawa et al., 2007; Nishikawa and Hori, 2015). As such, the physiological characteristics of E. zodiacus would have made it possible to have an advantage in the less affluent environment in March (low nitrogen and temperature conditions).

In April, similar to March, there were high abundances of Cryptomonas spp. and the diatom Skeletonema spp. Tidal currents in coastal waters play a dual role, in that they increase mixing of the water mass and they also increase translocation of phytoplankton (Shearman and Lentz, 2004; Queiroga et al., 2007; Shanks et al., 2014). Skeletonema spp. in particular is common in coastal and estuarine waters worldwide, except for polar regions (Kooistra et al., 2008; Assmy et al., 2019). This diatom can also adapt to brackish water, as demonstrated by Brand (1984) and Rijstenbil (1989). We found that the proportion of Skeletonema spp. was consistently high in the outer stations of the Seomjin River (S12), indicating that it may have been transported or dispersed by tidal currents following initial growth under the high nutrient and low salinity conditions near river mouths. Previous studies by Baek et al. (2015) and Lee et al. (2018) also showed that Skeletonema spp. accounted for a significant amount of the total phytoplankton biomass in the Gwangyang Bay and Seomjin River of Korea. Therefore, the results of our field observations and these previous studies suggest that Skeletonema spp. can proliferate under conditions of high nutrients and low salinity. In addition to Cryptomonas spp. and Skeletonema spp., Chaetoceros spp. and E. zodiacus were also highly abundant during our spring field survey. Our mesocosm experiments (described below) were designed to improve understanding of the temporal dynamics of these species following nutrient additions. Specifically, our mesocosm studies focused on the species that dominated during different stages: Cryptomonas spp., E. zodiacus, and Chaetoceros spp. during the early and middle stages, and C. closterium and Pseudo-nitzschia spp. during the middle and later stages.

The maximum phytoplankton abundance in the mesocosm experiment was on day 6 in the LN group and day 8 in the HN group, coinciding with timing of nutrient depletion. Notably, in our spring mesocosm study, the response of phytoplankton to nutrient infusion exhibited a slower growth compared to a recent summer mesocosm experiment (Yoon et al., 2023). In this previous summer mesocosm study, phytoplankton had a large rapid response to nutrient addition, and we attribute this to the higher water temperature and the higher level of nutrients during the summer. The nutrient uptake and cellular metabolism of phytoplankton were accelerated at high water temperatures (Montagnes et al., 2003; Kingsolver, 2009; Thomas et al., 2012). The mean water temperature of our spring mesocosms was 12.82 ± 0.62°C, but the previously described summer mesocosms (Yoon et al., 2023) had a mean water temperature of 28.54 ± 1.29°C, a difference of approximately 16°C. Thus, we attribute the two- to four-day delay in nutrient depletion and attainment of maximum phytoplankton biomass in our spring mesocosms to the lower water temperature. In other words, temperature affects nutrient consumption by phytoplankton, and a lower temperature during spring delays the proliferation of phytoplankton by several days.

At the beginning of our mesocosm experiment, the phytoplankton community was dominated by Cryptomonas spp. and E. zodiacus. Although the abundance of Cryptomonas spp. remained relatively stable in the control group throughout the experimental period, the abundances changed significantly in the LN and HN groups. As mentioned above, the diatom E. zodiacus may have been dominant in initial natural seawater (in study site) based on the physiological characteristics of relatively high nutrient uptake rate under low temperatures (Nishikawa et al., 2009). Based on the competitive advantage of high initial cell density (Grover, 1991), the cell density and biovolume of E. zodiacus rapidly increased with the use of abundant nutrients ( Figure 8 ), but the growth rate of E. zodiacus in LN and HN groups was already decreasing from day 2. On the day 6, when nutrients were not depleted, the dominant species shifted from E. zodiacus to Chaetoceros spp. The next dominant diatom Chaetoceros spp., characterized by rapid growth under high nutrient conditions (Lomas and Glibert, 2000; Shevchenko et al., 2004; Sinclair et al., 2009), and in this mesocosm study, the maximum growth rate was over 1.7 d^-1, which was substantially higher than that of E. zodiacus (0.65 d^-1) ( Table 1 ). This suggests that the fast and effective nutrient uptake and growth ability of Chaetoceros allowed them to dominate even at low initial cell densities. Consequently, we propose that E. zodiacus and Chaetoceros spp. have a high capacity for growth in nutrient-rich waters due to their efficient uptake of nutrients, as also demonstrated in previous studies (Estrada and Berdalet, 1997; Collos et al., 2005; Nishikawa and Hori, 2004; Tantanasarit et al., 2013). However, it suggests that not only the nutrient uptake capacity but also the initial cell density is highly important in determining dominance.

Following nutrient depletion in the mesocosm experiment, the phytoplankton C. closterium and Pseudo-nitzschia spp. gradually dominated. These two relatively small species have high surface-to-volume ratios, which allows for efficient nutrient uptake even in nutrient-depleted conditions. This suggests these small species have the ability to adapt and grow well under conditions of limited nutrients (Harrison and Hurd, 2001; Cochlan et al., 2008; Loureiro et al., 2009). Recent mesocosm studies by Park et al. (2020) and Yoon et al. (2023) also demonstrated the dominance of these two species in nutrient-depleted conditions. In agreement, our PCA analysis revealed significant negative correlations between the abundance of C. closterium and the levels of nitrate, phosphate, and silicate. The dominant species shifted to changes in nutrient levels, leading to changes in cell size as well as cell density. In mesocosm, which is a closed ecosystem and uses only limited resources, biovolume combining cell volume and cell density may be a more reasonable way to track nutrient use, especially by phytoplankton (Thomas et al., 2018). It revealed that the resource utilization of the dominant diatoms could underestimate the biovolume of E. zodiacus and overestimate the biovolume of Pseudo-niztschia spp. based on cell density ( Figure 8 ). On the other hand, in our mesocosm, this difference between cell density and biovolume was not large due to the similarity of cell size in dominant diatoms, which conversely proves that the analysis using cell density is reasonable. However, this can be completely reversed when macroplankton is introduced in the community, so the biovolume that can better explain the use of resources in the ecosystem needs to be considered in future studies.

In previous study (Smetacek, 1999; Silkin et al., 2013), in response to nutrient, a succession from cryptophytes to diatoms in field have reported, and this phenomenon was also reproduced in our mesocosm experiment. The reproducibility of these phenomena demonstrates the potential of mesocosms to represent the field and suggest that it can be used as powerful tools to bridge the large gap between laboratory-based experiments and field. Among our field survey, there was sampling site similar to the initial conditions of our mesocosm experiment (high nutrient enrichment and predominance of Cryptomonas spp. at S6). Although cautious, based on the results of our mesocosm experiments, it is possible to predict the dominance of diatoms specialized in nutrients in the future at that site. Studying phytoplankton community changes through mesocosm experiments, which replicate field conditions while controlling influential factors, aids in comprehending the causes of phytoplankton changes in natural environments and occasionally predicting field outcomes.