The Yellow Sea, consisting of wide intertidal flats and many islands, is a shallow semi-enclosed sea that has an average depth of 44 m with characteristic tidal amplitudes up to 10 m. The coastal regions of the Yellow Sea are affected by multiple environmental forces, with tide current variability being the most prominent (Hwang et al., 2014). In summer, strong stratification exists deep offshore due to solar heating, in contrast to tidal well-mixing in coastal regions (Kwon et al., 2011).

The HRE and YRE are parts of the Yellow Sea ( Figures 1A, B ) with mean spring tidal ranges of 5 and 8 m, respectively (Byun et al., 2004). Daily mean discharge rates range from 149 to 2326 m³ s⁻¹ for the HRE (Park et al., 2002) and from 0 to 285 m³ s⁻¹ for the YRE (Park and Sin, 2022). In the HRE, located near Incheon City, the tidal currents are predominantly flood currents toward the upper Han River through the southern channel of the Ganghwa Island during dry seasons, whereas ebb currents prevail toward the open sea through the Yeomha Channel during wet seasons (Yoon and Woo, 2013a; Lee et al., 2021). In the YRE, ebb currents dominate when the tidal flat changes into land, whereas this dominance disappears when the tidal flat changes into sea (Jung and Choi, 2010). In this study, we compared the compositions of phytoplankton in the HRE (in Incheon) and YRE (in Mokpo) by spatiotemporal variations.

Sampling and field observations were performed in November 2021 and February, May, and August 2022 to include four seasons: autumn (November), winter (February), spring (May), and summer (August). To conduct the sample collection in freshwater-affected area, five stations in the HRE (IC1–5) and four stations in the YRE (MP1–4) were determined based on the previous studies (Park et al., 2012; Yoon and Woo, 2015). In each sampling event, the water temperature and salinity were determined in situ using a YSI Model 85 probe (YSI Inc., Yellow Springs, OH, USA) and the Secchi and euphotic depths were obtained using a 30-m-diameter circular Secchi disk. A 6-l aliquot of seawater was collected from the sea surface and bottom using a Niskin sampling bottle at each site. The collected water samples were filtered through a 180-μm Nitex mesh net to remove large debris and zooplankton, then immediately transferred in to acid-washed polyethylene bottles and frozen at –20°C. The filter samples were stored at –20°C until photosynthetic analysis.

The water samples were thawed overnight at a low temperature (2°C) and brought to the laboratory prior to analysis. The concentrations of nitrate (NO₃ ⁻), nitrite (NO₂ ⁻), ammonium (NH₄ ⁺), silicate [Si(OH)₄], and phosphate (PO₄ ³⁻) were determined colorimetrically using a QuAAtro nutrient analyzer (SEAL Analytical GmbH, Norderstedt, Germany). The sum of NH₄ ⁺, NO₂ ⁻, and NO₃ ⁻ concentrations was considered as the DIN. PO₄ ³⁻ concentration was considered as the dissolved inorganic phosphorous (DIP), and the Si(OH)₄ concentrations was considered as the dissolved silicon (DSi). The measurements were performed in accordance with the Autoanalyzer’s practical manuals: NO₃ ⁻, NO₂ ⁻, and DSi (Hansen and Grasshoff, 1983), NH₄ ⁺ (Helder and De Vries, 1979), and DIP (Murphy and Riley, 1962).

To extract photosynthetic pigments, 47-mm Whatman GF/F filters were soaked in 95% methanol (5 ml) at −4°C for 12 h in the dark, then sonicated for 5 min to disrupt the cells, and centrifuged to remove filter debris and cells and aliquots (1 ml) were mixed with 300 μl of water before analysis. A ChroZen high-performance liquid chromatography (HPLC) system (Young-In Chromass, Ahn-yang, Korea) equipped with a Waters Symmetry C8 (4.6 mm × 150 mm; 3.5-μm particle size; 100-Å pore size) column (Waters, Milford, MA, United States) was used to determine photosynthetic pigments according to Zapata et al. (2000). The data matrix included concentrations of diagnostic biomarker pigments: peridinin, 19′-butanoyloxy-fucoxanthin, fucoxanthin, 19′-hexanoyloxy-fucoxanthin, neoxanthin, prasinoxanthin, violaxanthin, alloxanthin, lutein, zeaxanthin, Chl b (DHI, Denmark.), and Chl a (Sigma Co.) The concentration of standard pigments was determined using the absorption coefficients from Jeffery et al. (1997), and absorbance was calculated by adjusting absorbance values using the method described by Park and Park (1997). The pigments were quantified to analyze the community structure of phytoplankton using CHEMTAX by Mackey et al. (1996), based on the relative ratio of Chl a and pigments.

The abundance of individual phytoplankton groups was relative to total chlorophyll a (hereafter Chl a) in each water sample and was calculated using CHEMTAX (v. 1.95; https://cmr.earthdata.nasa.gov/search/concepts/C1214308429-AU_AADC) to obtain the phytoplankton composition from HPLC pigment data (Mackey et al., 1996; Wright et al., 1996). The steepest descent algorithm was used to optimize biomarker pigment-to-Chl a ratios to calculate the concentration of Chl a attributed to each class in the phytoplankton community. Previously published biomarker pigment-to-Chl a ratios for phytoplankton classes collected around the Korean peninsula were applied to initial pigment ratio matrices ( Supplementary Table 1 ) and corrected iteratively using CHEMTAX to minimize the sums of squares of differences between observed and calculated pigment concentrations (Lee et al., 2011). Eight different phytoplankton groups, namely, prasinophytes, chlorophytes, prymnesiophytes (typically coccolithophores), pelagophytes, dinoflagellates, cryptophytes, cyanobacteria, and diatoms, were uploaded for an initial CHEMTAX configuration. The data matrix included concentrations of diagnostic biomarker pigments: peridinin, 19′-butanoyloxy-fucoxanthin, fucoxanthin, 19′-hexanoyloxy-fucoxanthin, neoxanthin, prasinoxanthin, violaxanthin, alloxanthin, lutein, zeaxanthin, and Chl b. The final results were obtained in terms of the absolute concentrations (μg l⁻¹) of Chl a attributed to each phytoplankton class in a sample. Chl a concentrations estimated from CHEMTAX analysis were standardized (convert to percentile value) to calculate their relative portions, which were logarithmically transformed to allow the least abundant groups to contribute to the analysis.

From the percentile contribution of each phytoplankton obtained by CHEMTAX, we calculated phytoplankton group diversity and evenness using the Shannon–Weaver Diversity index (H′) and Pielou’s evenness index (j′), respectively (Pinckney et al., 2020).

Non-metric multidimensional scaling (nMDS) ordination was performed to unveil seasonal and spatial patterns in the phytoplankton community structure using the contributions of each phytoplankton to total Chl a obtained through CHEMTAX analysis. The percentile contributions of phytoplankton were transformed to log (X + 1) before the estimation of Bray–Curtis similarities (Bray and Curtis, 1957), and one-way analysis of similarity (ANOSIM) was used to analyze differences between phytoplankton classes due to spatiotemporal variations. The output of the test is the “Global R” (if Global R > 0, there is a separation between groups; if Global R < 0, there is no separation between groups). Similarity percentages (SIMPER) were used to identify classes that contributed to the differences in the composition of phytoplankton in each spatiotemporal class to support the ANOSIM results. PRIMER (v. 6.0) was used to conduct SIMPER and ANOSIM analyses. Prior to conducting statistical analyses, the normality of data was tested using the Shapiro–Wilk test and the homogeneity of variance was tested using Levene’s procedure. The Kruskal–Wallis nonparametric test was used to compare the seasonal and spatial distributions, followed by Dunn’s post hoc multiple pairwise comparison. ANOVA test was used to compare the spatial difference of phytoplankton alpha indices between the estuaries. These tests were conducted using IBM SPSS statistics software (v. 22.0, IBM Corp., Armonk, NY). Canonical correspondence analysis (CCA) was performed to visualize the relationships between the variability of the eight phytoplankton groups and biomass (expressed in terms of Chl a concentration) and physical and biogeochemical parameters (temperature, salinity, density, NO₃ ⁻, NH₄ ⁻, DSi, and DIP concentrations, and stoichiometric ratios among nitrogen, phosphorus, and silicate). A subsequent Monte Carlo permutation test was conducted to evaluate the statistical significance of the CCA results. CCA was performed using CANOCO statistical software (v. 4.5) (Ter Braak and Smilauer, 2002).

nMDS was performed based on the contribution of the eight phytoplankton groups to total Chl a, and the similarities of community structures were calculated and ordinated to visualize the spatiotemporal variations with the 0.17 stress ( Figure 2A ). The nMDS approach showed that the symbols for autumn observations in the HRE and YRE overlapped in the plot, whereas the symbols for winter observations were plotted with spatial differences. The symbols representing the observations in August 2022 were located without spatial differences between the two estuaries.

The ANOSIM test confirmed the seasonal and spatial dynamics of the phytoplankton communities in the HRE and YRE (HRE: Global R = 0.680, P = 0.001; YRE: Global R = 0.592, P = 0.001). Positive Global R values were obtained in both the seasonal and inter-estuarine comparisons during the study period ( Figure 2B ). The pairwise tests showed that phytoplankton communities in the HRE and YRE changed throughout the seasons (P < 0.01) (solid arrows in Figure 2B ). The Global R value was high in the autumn–winter transition in the HRE (Global R = 1.000, P < 0.001), whereas the highest value was obtained between spring and summer in the YRE (Global R = 0.475, P < 0.01). Each of the four seasons in the two estuaries was characterized by different phytoplankton structures (SIMPER results; Figure 2B ). Diatoms and cryptophytes were the phytoplankton groups contributing to the seasonal variation in the average dissimilarities of phytoplankton community structures in the HRE (>30% in the autumn and winter), whereas various phytoplankton groups contributed to the dissimilarities of phytoplankton community structures in the YRE. The highest and lowest values of average dissimilarity were observed in autumn–winter in the HRE and YRE, respectively.

The spatial comparison using the ANOSIM pairwise test and SIMPER analysis revealed that the phytoplankton community structures in the HRE and YRE were significantly different (dotted arrows in Figure 2B ). The inter-estuarine differences were remarkable during winter and summer with high Global R values (>0.5) and average dissimilarity (>30%). The SIMPER analysis revealed that diatoms (45.0%) and cryptophytes (36.1%) contributed the most to the highest average dissimilarity (47.4) in winter, while phytoplankton (diatoms, pelagophytes, and cryptophytes) contributed the least (9.9) in autumn. In spring, there were significant differences between the estuaries (P < 0.01) even though the phytoplankton groups were not separated (Global R < 0.5). The ANOSIM and SIMPER tests indicated that the seasonal succession of the phytoplankton community structure did not follow the coincident temporal pattern in the four seasons (P < 0.01).

The percentile contribution of the eight phytoplankton groups evidenced the differences between the continuously flushed HRE and the intermittently perturbed YRE, with diatoms being a highly dominant phytoplankton group contributing to the total Chl a in both HRE (60.3 ± 19.0%) and YRE (75.6 ± 17.8) ( Figure 3 ). Several seasonal differences were observed, including a higher contribution of cryptophytes from winter to summer (19.5 ± 16.0%) in the HRE and high contributions of pelagophytes, cyanobacteria, and prasinophytes with gradual increments toward the open sea in spring (pelagophytes: 17.4 ± 11.4%) and summer (cyanobacteria: 18.7 ± 14.4%; prasinophytes: 13.3 ± 7.0%) in the YRE.

During the study period, daily precipitation in neighboring cities ranged from 0 to 207.8 mm in Incheon City (for the HRE) and 0 to 81 mm in Mokpo City (for the YRE), with summer precipitation (June–August) being noticeably higher ( Figures 4A, B ). The daily discharge rate of the Singok submerged weir and the Yeongsan estuarine dam during the study period varied between 0 and 51.8 × 10³ m³ s⁻¹ for the HRE and 0 and 32.5 × 10³ m³ s⁻¹ for the YRE, with higher rates during the summer months (June–August) and highest frequencies in August ( Figures 4C, D ).

Seasonal variations in the seawater temperature exhibited typical patterns in the temperate coastal zone, with higher values in summer and lower values in winter (2.1°C–27.9°C in the HRE and 4.7°C–28.7°C in the YRE) ( Figures 4E, F ). Temperature and Salinity variations exhibited seasonal and spatial differences between the two estuaries ( Figures 4G, H ). In the HRE, the salinity gradient ranged from 1.2 to 31.4, with an obvious gradient depending on the distance from the Han River mouth in autumn, spring, and summer, whereas the salinity gradient was weak in winter. Distinguishably, the polyhaline zone was observed in the YRE, while the mesohaline zone was formed in the summer. The significant difference of salinity was observed in the HRE during the study period (Kruskal-Wallis, p < 0.05).

Temporal and spatial illustrations revealed variations in the distributions of dissolved inorganic nutrients ( Figures 5A–J ). During the study period, dissolved inorganic nitrogen (e.g. NO₃ ⁻, NO₂ ⁻, and NH₄ ⁺) were under replete and deplete conditions in the HRE and YRE, respectively. The HRE and YRE had no clear spatial differences in DIP and DSi levels.

NO₃ ⁻ was the dominant nitrogenous nutrient, with a strong relationship with the dissolved inorganic nitrogen (NO₃ ⁻ + NO₂ ⁻ + NH₄ ⁺ = DIN, r ^2 = 0.99, P < 0.001) concentration, exhibiting an obvious correlation with salinity (r ^2 = 0.61, P < 0.001) during the study period ( Figures 5A, B, E, F ). The NO₂ ⁻ concentration was negligible during the study period. There was no seasonal trend between the NO₃ ⁻ and DIN concentrations in the HRE (Kruskal–Wallis, df = 4, P = 0.418). The NH₄ ⁺ concentration had a significant seasonal variation, with lower values in spring in both HRE (Kruskal–Wallis, P = 0.003, df = 4) and YRE (Kruskal–Wallis, P = 0.000, df = 3), but no significant spatial variation in either estuary ( Figures 5C, D ).

The DSi concentration varied in the HRE, with lower values in spring (Kruskal–Wallis, P = 0.001, df = 4), displaying an obvious correlation with salinity (Spearman’s rank correlation analysis r = −0.53, P = 0.000) during the study period ( Figures 5G, H ). The DIP concentration in the water column exhibited seasonal differences in both estuaries, with lower values in spring (Kruskal–Wallis, P = 0.002, df = 4) in the HRE and lower values in winter and spring in the YRE (Kruskal–Wallis, P = 0.001, df = 3) ( Figures 5I, J ). In winter, significant differences in the DSi and DIP concentrations were observed between the HRE and YRE (Kruskal–Wallis, P < 0.001), with higher values in the HRE. A positive correlation between DIP and DSi was observed (Spearman’s rank correlation analysis, r = 0.909, P < 0.001) during the study period.

Depth-integrated Chl a ranged from 1.6 to 59.3 and 11.1 to 105.8 μg m⁻² in the HRE and YRE, respectively ( Figure 6 ). The significant difference of the depth-integrated Chl a between the HRE and the YRE was observed in winter (Mann-Whitney, p < 0.05). Biomass of phytoplankton based on the Chl a concentration indicated no significant seasonal variation in phytoplankton biomass in the HRE, whereas higher values were obtained in summer (Kruskal–Wallis, p = 0.03, df = 3).

The molecular N:P, Si:P, and Si:N ratios exhibited the stoichiometric limitation of nutrients, determined by the Redfield ratio (e.g., N:P:Si = 16:1:15) ( Figure 7 ). The P-limitation was noticeable in winter in both estuaries and in spring in the YRE ( Figure 7A ). There was no obvious N-limitation during the study period ( Figure 7B ). Si-limitation was observed in the outer station in the HRE (IC3–5) ( Figure 7C ).

CCA revealed the relationship between environmental parameters and the phytoplankton community structure and biomass in the HRE and YRE during the study period ( Figure 8 ). The first two axes (1 and 2) accounted for 49.6% and 32.7% of the total variance of the phytoplankton groups in relation to environmental factors in the HRE, respectively ( Figure 8A ). The CCA differentiated the cryptophytes as the group more associated with high DIN and NO₃ ⁻ concentrations. Small phytoplankton populations (e.g., chlrophytes, prasinophytes, pelagophytes, and cyanobacteria) were distributed in the same quadrant as summer symbols, with a positive correlation with the Si:N ratio and Chl a concentration and a negative correlation with the N:P ratio. The samples collected in spring differed in salinity and DIN concentration. In the YRE, with 50.7% for axis 1 and 33.1% for axis 2, explaining the total variance, the clockwise distribution of symbols follows the typical seasonality of temperature showing higher temperatures in summer and lower temperatures in winter ( Figure 8B ). There was an obvious distribution of the symbols in the spring samples with a positive relationship with stoichiometric ratios and a negative relationship with DIN.

The comparison of indices in phytoplankton diversity and evenness between two estuary revealed the spatial difference during the study period ( Table 1 ). The Shannon-Wiever Diversity Index (H′) ranged from 0.48 to 1.51 and from 0.24 to 1.47 in the HRE and YRE, respectively. Pielou’s evenness index (j′) varied from 0.23 to 0.97 for the HRE, and from 0.13 to 0.81 for the YRE, respectively. A significant difference of indices for diversity and eveness was observed between the two estuaries with a higher value in the HRE (one-way ANOVA, P = 0.02).

Notable characteristics of phytoplankton dynamics in this study were the contrasting seasonality of phytoplankton biomass and community structure between the continuously flushed estuary (the HRE) and the intermittently perturbed estuary (the YRE). Our results indicated no obvious seasonality of Chl a concentration in the HRE, contrasting with the summer maxima of Chl a in the YRE. The positive relationship between salinity and dissolved inorganic nutrients (e.g. NO₃ ⁻, DIN, DIP, and DSi) indicated that continuous river flow formed the eutrophic condition in the HRE. Specifically, our CCA supports the view that the standing stock of phytoplankton in the HRE was not largely responsible for the concentration of nutrients represented by the lower contribution of nitrogenous nutrients (relatively short arrows), showing no obvious relationship with Chl a (See Figure 8A ). Accordingly, we can hypothesize that no significant seasonality of phytoplankton biomass in the HRE was associated with the conservative characteristics of DIN, implying that the vulnerability of the ecosystem was regulated by leading eutrophication by heavy precipitation through freshwater runoff into the macrotidal estuary (Jahan and Choi, 2014). The record-high precipitation in August in adjacent cities (e.g., Seoul and Incheon) may form the polyhaline zone in the upper HRE ( Figure 4D ). In contrast to the HRE, the summer bloom of phytoplankton was conspicuously characterized by co-dominance of diatoms, cyanobacteria (station after the dam), and prasinophytes (outer zone). This summer bloom was consistent with prior findings (Sin et al., 2015; Kim and Shin, 2021; Park and Sin, 2022). The close relationship between Chl a and temperature in the CCA plot ( Figure 8B ), separated from higher DIN concentrations, indicates that intensive discharge in summer caused the mismatch between the supply and consumption of nutrients in the phytoplankton community.

In autumn, the two estuaries are concurrently characterized by the overwhelming dominance of diatoms (>70%). Our ANOSIM and SIMPER analyses revealed subtle (average dissimilarity = 18.0) but significant differences (Global R = 0.759) that cannot be explained by the overall differences in phytoplankton composition between the two estuaries due to the high contribution of diatoms ( Figure 2 ). The diatom dominance was explained by the seasonal cycle of abiotic conditions on the western coast of Korea, which included enhanced mixing in the water column due to tidal activity and seasonally strengthened winds, an adequate temperature range (15°C–20°C) (Kang et al., 2019), and thus replete conditions of DIN (see Figures 5A–E ). Considering the tidal flat as a source of inorganic nutrients, NH₄ ⁺ can be replenished by re-mineralization during summer and autumn and then exported offshore through active tidal mixing (Kim et al., 2011). High concentrations of NH₄ ⁺ and DIP were observed in autumn and summer during the study period. Nitrogen from nonpoint sources may also be utilized by phytoplankton flowing from adjacent cities (e.g., Incheon Harbor for the HRE and Mokpo Harbor for the YRE). This is supported by the lack of a close relationship between NH₄ ⁺ and salinity in the CCA plot (see Figures 8A, B ). In an urbanized estuary, NH₄ ⁺ can remain at concentrations sufficient to inhibit NO₃ ⁻ uptake when the flow is sufficiently high in the channel, stimulating diatom dominance (Dugdale et al., 2012).

Notably, conspicuous dissimilarities of phytoplankton community structure were observed in winter. Specifically, the HRE manifested remarkable differences in both seasonal (e.g., autumn and spring) and inter-estuarine comparisons, which were characterized by increasing diatoms and cryptophytes. During winter, dry conditions related to low precipitation minimize riverine flushing, thereby strengthening the effect of seawater intrusion through tidal mixing in the HRE (Lim et al., 2007; Lee et al., 2021). In addition, strong wind and tidal currents may form highly turbid conditions by resuspension of sediments, resulting in low light availability in the HRE (Park, 2007). Given seawater-dominant conditions, wind-driven current from the YS can supply NO₃ ^– via the Korean Coastal Current, magnifying the P-limitation in the HRE (Tak et al., 2022). Consequently, cryptophytes in the HRE can thrive in winter due to low temperatures, dry atmosphere, and current-driven P-limitation. In contrast to the HRE, the dominance of diatoms lasts from autumn to winter with no obvious change in the phytoplankton community structure in the YRE.

Our study reveals that seasonal changes in the phytoplankton community in the YRE are caused by the contributions of nano- and pico-plankton (e.g., prasinophytes, pelagophytes, prymnesiophytes, and cyanobacteria with Global R > 0.6). This finding is consistent with the community structure of the offshore area in the East/Japan Sea and the Yellow Sea adjacent to the Korea Peninsula. In spring, the abundance of pelagophytes in the YRE differentiated the seasonal and inter-estuarine distributions of the community structure, whereas the seasonal variation in the HRE was minimal. Pelagophytes are generally considered an oceanic group, and significant contributions of pelagophytes are observed in the Bay of Bengal (Pujari et al., 2019), the Bohai Bay (Wang et al., 2021), and the Korea Peninsula (Kim et al., 2010; Jang et al., 2021). These studies reported negative correlations between pelagophytes and NO₃ ^–, DSi, and diatoms, which is consistent with our results. A strong tide may deliver offshore water into the two estuaries, and a continuous inflow of freshwater replenishes nutrients in the HRE in contrast to stagnant conditions downstream of the dam in the YRE. In turn, continuous flushing alleviates nutrient limitations, resulting in consecutive phytoplankton community structures, evidenced by the co-occurrence of diatoms and cryptophytes, whereas depletion of nutrients in the stable water column results in the active growth of pelagophytes, tidally transported from offshore. Consequently, the difference in phytoplankton between the two estuaries is attributable to the presence and absence of artificial construction.

In summer, N-limitation was evident in the HRE even though the water column was in a eutrophic condition in terms of the concentration of nutrients. The N-limitation in downstream of continuously flushed by the river was in accordance with previous study conducted in the Seomjin River estuary located in the southern coast of Korea (Lee et al., 2018). Summer bloom in the YRE was followed by an increase in the amount of cyanobacteria and prasinophytes in summer. Notably, increasing nutrient availability was not a limiting factor for phytoplankton growth in the HRE but may influence the growth of phytoplankton during the summer discharging period in the YRE. The contribution of cyanobacteria in the HRE was significant (>20%) after the estuarine dam (MP1) and gradually decreased toward the open sea during summer, whereas that in the HRE was negligible. Cyanobacteria are competitive within the temperature range of 25°C–35°C (Lürling et al., 2013). Previous studies have reported that cyanobacteria can be a dominant phytoplankton group upstream of estuarine dams on the western coast of Korea because of the increased retention time, resulting in nutrient depletion (Sin et al., 2015; Kim et al., 2021b). Given that there is no significant difference in seawater temperatures between the two estuaries, cyanobacteria flourished in the YRE during summer because of a combination of optimal growth temperature and physical input through dam discharge water. A previous study demonstrated that opening the middle stream weir and estuarine dam in the Yeongsan River reduced the magnitude of summer cyanobacteria blooms by regulating the reservoir retention time (Kim and Shin, 2021). Furthermore, the opening of the dam and weir in the Yeongsan River reduced the magnitude and frequency of algal blooms, which is similar to the lack of obvious seasonality of the Chl a concentration in this study (Kim and Shin, 2021). This suggests that the estuarine dam influences the phytoplankton community structure by allowing drifting cyanobacterial to enter the dam through the sluice gate (Doi et al., 2008).

During the study period, the two estuaries were characterized by diatom dominance and phytoplankton communities distinguished by small phytoplankton populations (e.g., cryptophytes, prasinophytes, pelagophytes, and cyanobacteria), inducing the inter-estuarine dissimilarity of the phytoplankton community. Ubiquitous characteristics of diatoms are consistently observed in estuaries in the temperate zone, such as the marginal zone of the Yellow Sea (Zhou et al., 2008; Kim et al., 2023). Diatoms were the most prevalent group, owing to the well-mixed condition of shallow estuaries in the temperate zone (Mallin et al., 1991; Trigueros and Orive, 2001; Haese et al., 2007). Indeed, the increases in contributions of diatoms and cryptophytes were observed in the estuaries in the Nakdong River and the Geum River, which are major rivers in terms of dranage in Korea (Baek et al., 2019; Kim et al., 2019; Kim et al., 2023). This study showed the prevalence of diatoms in community composition (> 40%) throughout the year on the western coast of Korea. The results are consistent with previous reports, implying the dominance of diatoms under well-mixed conditions due to tidal mixing on the western coast of Korea (Park et al., 2022). Despite the limitation of the CHEMTAX method for depicting the phytoplankton community structure, which can only provide groups at the class level, the negative relationship between diatom contribution (%) and α-indices (e.g., Shannon–Wiener diversity index and Pielou evenness index) indicates that the regulation of the community structure on the western coast of Korea is strongly associated with the phenology of diatoms ( Supplementary Figures 1A, B ).

Cryptophytes were the second most prevalent group among the phytoplankton community structures in the HRE. Previous studies have revealed a positive correlation between cryptophytes and nitrate loading, resulting from P-limitation in estuarine areas in Korea (Bibi et al., 2020) and the eastern coast (Kang et al., 2021), characterized by continuous flushing of riverine input. Our CCA results explained the thriving of cryptophytes in the HRE with nitrate-associated DIN concentrations.

The swing of diatoms and cryptophytes was evident in the study area ( Figure 8 and Supplementary Figure 1C ). The negative correlation between these two groups is evidenced in our CCA results in both the HRE and YRE ( Figures 8A, B ). This result is consistent with previous reports on temperate estuaries (Kang et al., 2021; Santos et al., 2022). Several studies have presented reasons for the increase in cryptophyte dominance in estuaries with P-limitation (Hyun et al., 2011; Santos et al., 2022), such as seasonal variations in environmental conditions, grazing impact of shellfish, and climate change (Jahan and Choi, 2014). For example, diatom–small flagellate alternation is typical on the Mediterranean coasts, characterized by the dominance of cryptophytes associated with well-mixed conditions under lower light availability during winter (Zingone et al., 2010; D’Alelio et al., 2015). Given that the HRE has a higher concentration of total suspended solids (TSS) than the YRE, which is governed by freshwater runoff (Park, 2007), the enhanced contribution of cryptophytes during winter supports the contrast structure between the HRE and YRE in terms of the phenology of cryptophytes ( Figures 3C, D ). In addition to the thriving of cryptophytes, the recently declining concentration of nitrogenous nutrients but increasing N:P ratio by water quality control, referred to as “P-only strategy,” presumably leads to a dramatic swing of diatoms and cryptophytes, reflecting the DIN regulation of the phytoplankton community structure in long-run (Kim et al., 2021a).

The western coast of Korea has recently experienced changes in physicochemical conditions such as winter warming of the surface seawater (Jahan and Choi, 2014; Park et al., 2015) and an increase/decrease in the amount of dissolved nutrients (Park et al., 2023). The response of macrotidal estuaries to nitrogen loading differs, which is mainly regulated by sediment resuspension, and thus, increasing turbidity hinders the active consumption of nutrients (Monbet, 1992). Kim et al. (2017) reported similar seasonality between Chl a concentration and TSS in the Gyeonggi Bay (the HRE) and the southwestern coast of Korea (adjacent to the YRE), which are profoundly affected by tidal currents. As these coastal areas are characterized by shallow continental shelves, the tide-dominant environment is vulnerable to changes in physical conditions, which can cause different responses of the phytoplankton community structure by determining the magnitude of energy transport from organic matter derived from phytoplankton along the coastal processes (Najjar et al., 2010; Hwang et al., 2014; Lee et al., 2022). Our study supports the vulnerability of the phytoplankton community structure in shallow macrotidal estuaries due to increased differences during winter combined with macrotidal conditions, strengthened monsoonal winds, and minimized freshwater input. Considering this evident difference in the phytoplankton community structure in winter, climatic changes in winter (e.g., increasing seawater temperature and precipitation) can cause unpredictable changes in the ecosystem structure (Najjar et al., 2010). Along with climatic changes, long-term variations in the nutrient concentration and ratios may influence the phytoplankton community structure, for example, increasing N:P ratio and declining total nitrogen downstream in the Han River, in contrast to the increasing DIN concentration in the Yeongsan River. Pinckney et al. (2020) reported 25 μM as the DIN breakpoint for phytoplankton group diversity in high-saline coastal water (22–36 salinity), with the following unpredictability above this breakpoint. Despite spatial variability in DIN breakpoints, the long-term trend of increasing DIN suggests that the eutrophication of coastal areas in Korea may cause a shift in the phytoplankton community structure.

Damming of rivers is considered a cataclysmic event in the riverine ecosystem because it disconnects habitats from secondary and primary producers in the coastal ecosystem such as filter-feeding bivalve and fish larvae (Ligon et al., 1995; Lee et al., 2022). Flushing serves as a regulator of phytoplankton and its consumer in lower estuaries by controlling salinity, light availability, and material retention time and/or by enhancing/reducing nutrient limitations (Schöne et al., 2003; Murrell et al., 2007). The case of the Three Gorges Dam, which is connected to the East China Sea, the world’s largest hydroelectric construction, may well explain the impact of dams, implying the alternation of estuaries with high variability in the microbial structure depending on dam discharge (Jiao et al., 2007). Our study shows the flourishing of cyanobacteria near the dam in summer, presumably transported from dam discharge. Considering the possible contribution of reservoir plankton to nitrogen uptake by consumers as a form of fine particulate organic matter, cyanobacteria can supply trophic subsidy to the coastal food web through dam discharge (Doi et al., 2008; Ock and Takemon, 2014). The pelagic food web structure can be shifted by changing the dominant phytoplankton (diatoms–cryptophytes–cyanobacteria), which can be triggered by variability in dominant forms and nutrient amounts and ratios through the phytoplankton community structure (Glibert, 2010). Cryptophytes have a low biovolume; thus, increasing their dominance may be insufficient to compensate for the loss of diatom contribution in terms of energy transfers from primary producers to higher-trophic-level organisms (Lehman et al., 2010). Therefore, the presence of an estuarine dam can cause a shift in the downstream ecosystem via bottom-up regulation of phytoplankton initiated by the discontinuity of riverine–estuarine–coastal continuum, whereas the contribution of small-size phytoplankton is enhanced by water quality treatment of nutrients, intensifying P-limitation on the western coast of Korea.