Eutrophication Driven by Aquaculture Fish Farms Controls Phytoplankton and Dinoflagellate Cyst Abundance in the Southern Coastal Waters of Korea

We examined the dynamics of dinoflagellate cyst and phytoplankton assemblages in eutrophic coastal waters of Korea, adjacent to fish and shellfish farms. Water temperature showed seasonality, whereas salinity and pH remained relatively consistent. Dissolved inorganic nutrient levels were higher in September and at the inner stations, where aquaculture fish farms are located than those in May and at the outer stations. Canonical correspondence analysis and artificial neural network analysis revealed multiple environmental factors that affect the distribution of phytoplankton and dinoflagellate cysts. Diatoms dominated in the phytoplankton assemblages, while the protoperidinioid group dominated in the dinoflagellate cyst assemblages. Cyst abundance was higher at the outer stations than at the inner stations due to transport by fast currents, and phytoplankton abundance was positively correlated with cyst abundance. An increase in diatom abundance led to an increase in heterotrophic/mixotrophic cyst abundance, indicating that excessive uneaten food and urinary waste from the fish farms caused eutrophication in the study region and fast growth of diatoms, thereby contributing to the growth of heterotrophic/mixotrophic dinoflagellates and consequently, high abundance of heterotrophic/mixotrophic dinoflagellate cysts.


Introduction
Dinoflagellates are eukaryotic algae that produce resting cysts via sexual reproduction and/or temporary cysts in response to changes in environmental conditions such as temperature, salinity, and light [1]. The resting stage is a part of their life cycle, and the accumulated cysts can often resist harsh environmental conditions in sediments [2][3][4]; resting cysts play a pivotal role as seed populations, affecting phytoplankton communities during phytoplankton blooms [2,[5][6][7]. Monitoring dinoflagellate cysts enables the temporal and spatial prediction of bloom initiation [2,8], tracking of historical records of harmful algal blooms [9], and analogy of historic environments in the water column and sediments [10].
As phytoplankton community structure varies with changes in environmental conditions and geographical characteristics of the study region and both species composition and abundance explicitly respond to physical and chemical conditions in the water column, phytoplankton are important for characterizing seawater environments [11]. For this reason, numerous previous studies have investigated how dinoflagellate cysts are associated with vegetative cells using environmental variables in coastal countries across the globe, including those in Asia [12][13][14][15][16][17], Europe [18][19][20][21], and North America [22][23][24].
Nutrients introduced from land via runoff contribute to high primary productivity in coastal waters, but excessive nutrient input exacerbates water quality and often causes harmful algal blooms [25,26]. Massive aquaculture farms are located along the southern

Field Samples
Sampling was performed at five stations in two days in May and September 2006 with three inner stations (Stations 1, 2, and 3) and two outer stations (Stations 4 and 5; Figure 1). We collected samples for biotic (dinoflagellate cysts and phytoplankton) and abiotic samples (nutrients and sediment environments). Seawater samples were collected

Field Samples
Sampling was performed at five stations in two days in May and September 2006 with three inner stations (Stations 1, 2, and 3) and two outer stations (Stations 4 and 5; Figure 1). We collected samples for biotic (dinoflagellate cysts and phytoplankton) and abiotic samples (nutrients and sediment environments). Seawater samples were collected using a Niskin water sampler (General Oceanics, Miami, FL, USA) at 1 m below the water surface and 1 m above the bottom sediment; 1 L samples were fixed in polyethylene bottles with Lugol's solution at a final concentration of 1%. The samples were covered with aluminum foil to inhibit fixative degradation by sunlight and were kept in a cooler with ice until the samples were delivered to the laboratory. Surface sediment samples were collected using a TFO gravity corer (University of Tokyo, Fisheries Oceanography Laboratory), consisting of a 30 cm-long and 1 cm-diameter tube. These tubes with 30 cmlong sediments were covered with aluminum foil to inhibit cyst germination by light and stored at 4 • C until further analyses. Subsequently, 20 mL nutrient samples were collected by filtering onto pre-combusted (2 h at 250 • C) glass fiber filters and stored in a −20 • C freezer until analysis. Physical and chemical parameters at the surface and bottom waters, including temperature, salinity, pH, and dissolved oxygen (DO), were measured on-site using a YSI 556 (YSI Inc., Yellow Springs, OH, USA).

Water and Sediment Sample Analysis for Environmental Variables
Ammonium, nitrate, nitrite, and phosphate were analyzed in duplicate using standard spectrophotometric methods [36][37][38]. Dissolved inorganic nitrogen (DIN) is defined as the summation of ammonium, nitrate, and nitrite. Water content was determined using the top 3 cm of sediments by calculating the ratio of the weight difference between the wet sediment and dried sediment (24 h at 110 • C) relative to the initial weight of the wet sediment. Sediment samples that were utilized to measure water content were also used to determine ignition loss (IL) by comparing the weight difference in the samples before and after combusting for 4 h at 550 • C. Generally, a direct measurement such as IL or organic carbon is utilized to estimate organic matter content in sediments [39], while chemical oxygen demand (COD) is utilized to quantify the amount of consumed oxygen during oxidation of organic matter [40]. COD was measured using the alkaline potassium permanganate method [41]. Acid volatile sulfide (AVS) is defined as the amount of hydrogen sulfide generated under anoxic conditions and is measured by converting sulfide in sediment to hydrogen sulfide [42]. To determine AVS, 2 g of sediment was transferred to a gas-generating tube and 2 mL of sulfuric acid (18N H 2 SO 4 ) was added to measure the hydrogen sulfide using a gas detection tube.
The eutrophication index (E) is calculated to determine the degree of eutrophication in the study region. The index was calculated using COD and nutrients via derivation from a globally and locally applied equation, E = (COD × DIN × DIP)/3.43, where E is the eutrophication index, COD and DIN are as defined above, DIP is dissolved inorganic phosphate. When E > 1, the region is eutrophic and when E < 1, the region is not eutrophic [43,44].

Phytoplankton and Dinoflagellate Cyst Assemblage Analysis
Phytoplankton samples were transferred to settling tubes and settled for 48 h in the dark. The supernatant was then removed to yield 10 mL of concentrated samples [45]. Then, 1 mL of the final sample was mounted on a Sedgewick-Rafter counting chamber to quantify the phytoplankton assemblages using a light microscope (Olympus CH30; Olympus Corporation, Tokyo, Japan). Phytoplankton identification followed Shim [46] and Tomas [47].
Sample analysis for dinoflagellate cysts followed the paleontological method outlined in Matsuoka and Fukuyo [4]. The top 3 cm of the 30 cm-long sediment samples were weighed and then stored in the dark for 24 h after adding 15 mL of 10% hydrochloric acid to remove calcareous matter (e.g., foraminifera and fraction of shell). The acid-amended samples were washed multiple times using distilled water and stored in the dark after adding 15 mL of 47% fluoride acid to dissolve any siliceous matter (e.g., sand and diatom frustule). The samples were washed multiple times until the pH was 7 (i.e., samples turned neutral) and then transferred to 100 mL glass beakers to form a slurry. The samples were sonicated for 30 s and sieved using 125 µm and 20 µm sieves. Residual samples were transferred to 20 mL polyethylene tubes to create the final samples for cyst quantification. Finally, 1 mL samples were mounted on a Sedgewick-Rafter chamber to quantify the dinoflagellate cysts using an inverted microscope (AXIOVERT 200; Carl Zeiss AG, Oberkochen, Germany). The abundance of the dinoflagellate cysts was presented in terms of cysts/g dry as follows: where N: total abundance of dinoflagellate cysts, W: weight of wet sediment (g), and R: ratio of sediment water content. N was obtained by multiplying the number of the counted cyst by 20, so that the number of cysts in 20 mL was calculated. Cyst identification followed Bolch and Hallegraeff [48], Nehring [8], and Matsuoka and Fukuyo [4].

Data Analysis and Statistical Analysis
To test the significant differences in the environmental variables between the surface water and bottom water and among stations, a Student t-test was performed on the water quality and sediment environmental variables, while the Kruskal-Wallis test was conducted to compare the DIN and DIP levels among the southern coastal waters of Korea. A Wilcoxon rank-sum test was utilized to compare the differences in the abundance of the biotic variables between May and September. A linear regression analysis was conducted on the cell abundance between the dinoflagellate cysts and phytoplankton to assess the relationship between two biological parameters. A canonical correspondence analysis (CCA) was performed to elucidate the relationship between the physicochemical variables of the seawater and phytoplankton assemblages and the environmental variables of the sediment and dinoflagellate cyst assemblages [27]. Statistical information of CCA from the phytoplankton and dinoflagellate cyst assemblages are presented in Tables S1 and S2, and the scree plot of each community is exhibited in Figures S1 and S2, respectively. An artificial neural network (ANN) model was assembled to evaluate the environmental variables with the most influence on the phytoplankton and dinoflagellate cyst abundance as a function of the water environmental variables and sediment environmental variables, respectively. A fundamental objective of recent ANN analysis tools including neuralnet [49], nnet [50], and RSNNS [51] is to address the concern that supervised neural networks are "black boxes" that provide no sufficient information about underlying relationships between variables [52,53]. The most popular form of neural network is the feed-forward multilayer perception trained using an error backpropagation training algorithm. The backpropagated error computed between the observed and estimated results is utilized to adjust the connection weights. This minimizes the error between the desired and predicted outputs [54,55]. The input data were transformed to a log form because the ranges of input data were wide. After the log-transformation, the dataset was scaled to a range from 0 to 1, prior to a train. ANN was operated predicting from weights and output data, while the modeled value was fed forward and compared to the measured response, from which the mean square error (MSE) was computed as 9.08 for phytoplankton and 6.84 for dinoflagellate cysts. 70% of data was utilized for training and 30% of data was utilized for validation. Measured environmental variables were considered as input neurons for ANN modeling processes, including temperature, salinity, pH, DO, DIN, and DIP for phytoplankton, and IL, AVS, COD, and water content for dinoflagellate cysts. ANN is usually applied to predict the response of one or more variables against one to many explanatory variables [56]. Thus, ANN is often used to elucidate the relative strength of environmental forces shaping phytoplankton biomass and community composition as a function of environmental variables [57][58][59][60]. A comparison test for the mean including a Student t-test, Kruskal-Wallis test, and Wilcoxon rank-sum test was performed using R (R Foundation for Statistical Computing, Vienna, Austria). The CCA was executed using XLSTAT (Addinsoft, Paris, France), and the ANN was accomplished using the package 'neuralnet' in R.

Environmental Conditions in the Water Column
The environmental variables of the seawater slightly changed across the sampling stations ( Figure 2). The temperature varied in the range 12.  Figure 2B). Salinity was fairly stable in May with 33.93 ± 0.08 psu at the surface and 33.90 ± 0.05 psu at the bottom ( Figure 2C), whereas it was slightly lower in September at 31.13 ± 0.03 psu in the surface water and 31.33 ± 0.17 psu in the bottom water ( Figure 2D). In September, the bottom salinity was slightly higher at the outer stations (Stations 4 and 5) than that at the inner stations (31.53 and 31.48 psu, respectively; Figure 2D). pH was not significantly different between the surface and bottom waters in May and September (p > 0.05; Student t-test), except at Station 2 in May (p < 0.05; Student t-test). However, pH at the inner stations, which were located close to the aquaculture fish farms, was slightly lower than that at the outer stations ( Figure 2E,F). In May, DO in the bottom water (8.38 ± 0.13 mg/L) was significantly higher than that in the surface water (7.87 ± 0.39 mg/L; p < 0.05; Student t-test), except at 5 ( Figure 2G). However, in September, the DO was fairly consistent between the surface and bottom waters at 3.26 ± 0.51 mg/L and 3.13 ± 0.30 mg/L, respectively ( Figure 2H). DO at the outer stations was significantly higher than that at the inner stations (p < 0.05; Student t-test; Figure 2G,H).
Generally, dissolved inorganic nutrient levels were lower at the outer stations than at the inner stations, while the differences between the surface and bottom waters in September were significantly higher (p < 0.05; Kruskal-Wallis test) compared to those in May ( Figure 3). Ammonium levels varied in the range 3.07-4.49 µM (3.68 ± 0.55 µM) in the surface water, with 3.44-4.09 µM (3.69 ± 0.25 µM) in the bottom water in May ( Figure 3A), but the levels were slightly lower in September, with a significant reduction at the outer stations (1.65 ± 0.02 µM) compared to those at the inner stations (3.77 ± 0.31 µM; p < 0.05; Kruskal-Wallis test; Figure 3B). Nitrite levels were relatively consistent in May and September ( Figure 3C,D), but nitrate levels in the surface water and bottom water significantly increased from 0.51 ± 0.03 µM and 0.48 ± 0.57 µM in May to 3.99 ± 0.31 µM and 4.84 ± 0.86 µM in September, respectively (p < 0.05; Kruskal-Wallis test; Figure 3E,F). While the DIN levels in May were fairly stable among all stations, varying in the rage of 4.64-6.12 µM (5.30 ± 0.57 µM) in the surface water and 5.00-5.83 µM (5.34 ± 0.32 µM) in the bottom water ( Figure 3G), the DIN levels in September significantly decreased from 9.09 ± 1.20 µM in the surface water and 8.54 ± 0.22 µM in the bottom water at the inner stations to 6.70 ± 0.41 µM in the surface water and 5.88 ± 0.04 µM in the bottom water at the outer stations (p < 0.05; Kruskal-Wallis test; Figure 3H). The DIP levels in May slightly decreased toward the outer stations with 0.79 ± 0.12 µM in the surface water and 0.92 ± 0.09 µM in the bottom water ( Figure 3I). The decreasing pattern was similar in September, but the levels were moderately high with 1.33 ± 0.22 µM in the surface water and 1.55 ± 0.05 µM in the bottom water ( Figure 3J). Generally, dissolved inorganic nutrient levels were lower at the outer stations than at the inner stations, while the differences between the surface and bottom waters in September were significantly higher (p < 0.05; Kruskal-Wallis test) compared to those in May ( Figure 3). Ammonium levels varied in the range 3.07-4.49 µM (3.68 ± 0.55 µM) in the surface water, with 3.44-4.09 µM (3.69 ± 0.25 µM) in the bottom water in May ( Figure 3A), but the levels were slightly lower in September, with a significant reduction at the outer stations (1.65 ± 0.02 µM) compared to those at the inner stations (3.77 ± 0.31 µM; p < 0.05; Kruskal-Wallis test; Figure 3B). Nitrite levels were relatively consistent in May and September ( Figure 3C,D), but nitrate levels in the surface water and bottom water significantly increased from 0.51 ± 0.03 µM and 0.48 ± 0.57 µM in May to 3.99 ± 0.31 µM and 4.84 ± 0.86 µM in September, respectively (p < 0.05; Kruskal-Wallis test; Figure 3E,F). While the 6.70 ± 0.41 µM in the surface water and 5.88 ± 0.04 µM in the bottom water at the outer stations (p < 0.05; Kruskal-Wallis test; Figure 3H). The DIP levels in May slightly decreased toward the outer stations with 0.79 ± 0.12 µM in the surface water and 0.92 ± 0.09 µM in the bottom water ( Figure 3I). The decreasing pattern was similar in September, but the levels were moderately high with 1.33 ± 0.22 µM in the surface water and 1.55 ± 0.05 µM in the bottom water ( Figure 3J).

Environmental Conditions in Sediments
IL (%), AVS (mg/g dry), COD (mg O 2 /g), and water content (%) were measured to assess the environmental variation in the surface sediments ( Figure 4). IL was stable at 7.07-8.11% (7.48 ± 0.41%) in May and 6.57-7.17% (6.88 ± 0.24%) in September ( Figure 4A,B), whereas AVS sharply increased toward the outer stations with levels of 0.11 ± 0.03 mg/g dry at the inner stations and 0.17 ± 0.11 mg/g dry at the outer stations in May and 0.05 ± 0.00 mg/g dry at the inner stations and 0.07 ± 0.02 mg/g dry at the outer stations in September ( Figure 4C,D). COD was not significantly different between May (21.16 ± 1.97 mg O 2 /g) and September (21.97 ± 1.96 mg O 2 /g; p > 0.05; Student t-test; Figure

Environmental Conditions in Sediments
IL (%), AVS (mg/g dry), COD (mg O2/g), and water content (%) were measured to assess the environmental variation in the surface sediments ( Figure 4). IL was stable at 7.07-8.11% (7.48 ± 0.41%) in May and 6.57-7.17% (6.88 ± 0.24%) in September ( Figure  4A,B), whereas AVS sharply increased toward the outer stations with levels of 0.11 ± 0.03 mg/g dry at the inner stations and 0.17 ± 0.11 mg/g dry at the outer stations in May and 0.05 ± 0.00 mg/g dry at the inner stations and 0.07 ± 0.02 mg/g dry at the outer stations in September ( Figure 4C,D). COD was not significantly different between May (21.16 ± 1.97 mg O2/g) and September (21.97 ± 1.96 mg O2/g; p > 0.05; Student t-test; Figure 4E

Species\Stations
May September   Dinoflagellate cysts were also quantified as either autotrophic or heterotrophic/mixotrophic species by trophic strategy (Table 3; Figure 6). In May, the abundance of auto- Dinoflagellate cysts were also quantified as either autotrophic or heterotrophic/mixotrophic species by trophic strategy (Table 3; Figure 6). In May, the abundance of autotrophic species ranged from 1080 to 2460 cysts/g (mean of 1820 cysts/g) with 38% dominance and the abundance of heterotrophic/mixotrophic species ranged from 1940 to 4900 cysts/g (2980 cysts/g) with 62% dominance (Table 3; Figure 6A). The relative abundance of heterotrophic/mixotrophic species was higher at Stations 4 and 5 than at the inner stations ( Figure 6A). In September, the abundance of autotrophic species was between 1500 and 4360 cysts/g (2660 cysts/g), accounting for 42% of the total abundance, while the abundance of heterotrophic/mixotrophic species was between 2720 and 5380 cysts/g (3680 cysts/g) with 58% dominance (Table 3; Figure 6B). The relative abundance of heterotrophic/mixotrophic species at the outer stations was still higher (>50%) compared to that at the inner stations ( Figure 6B). The ratio of autotrophic species to heterotrophic/mixotrophic species slightly decreased in September because more autotrophic species were identified, including the genus Scrippsiella in the calciodineloid group and the genus Alexandrium in the gonyaulacoid group (Table 3).

Paleontological Name
ar. Sci. Eng. 2021, 9,362 15 o trophic species ranged from 1080 to 2460 cysts/g (mean of 1820 cysts/g) with 38% do nance and the abundance of heterotrophic/mixotrophic species ranged from 1940 to 4 cysts/g (2980 cysts/g) with 62% dominance (Tables 3; Figure 6A). The relative abunda of heterotrophic/mixotrophic species was higher at Stations 4 and 5 than at the inner s tions ( Figure 6A). In September, the abundance of autotrophic species was between 1 and 4360 cysts/g (2660 cysts/g), accounting for 42% of the total abundance, while the ab dance of heterotrophic/mixotrophic species was between 2720 and 5380 cysts/g (3 cysts/g) with 58% dominance (Table 3; Figure 6B). The relative abundance of hete trophic/mixotrophic species at the outer stations was still higher (>50%) compared to t at the inner stations ( Figure 6B). The ratio of autotrophic species to heterotrophic/mi trophic species slightly decreased in September because more autotrophic species w identified, including the genus Scrippsiella in the calciodineloid group and the genus exandrium in the gonyaulacoid group (Table 3).

Relationship between Environmental Variables and Biotic Variables
The CCA elucidated the relationship between the environmental variables and bio variables (Figure 7). Phytoplankton assemblages were distinctively clustered by temp ature and salinity, generating a clear segregation between the May and September pop lations ( Figure 7A). The September phytoplankton communities exhibited a positive c relation with temperature and DIP, while the May communities presented a positive c relation with salinity and DO ( Figure 7A). Particularly, diatoms and dinoflagellates w positively correlated with temperature and DIP ( Figure 7A). In contrast to the phytopla ton assemblages, the CCA results for the dinoflagellate cyst assemblages showed no cl clusters between the two seasons, while COD and AVS were the most influential envir mental variables on the dinoflagellate cyst communities ( Figure 7B). The CCA revea that the gonyaulacoid group was positively correlated with COD and AVS, while the p toperidinioid group was negatively correlated with COD and AVS but positively cor lated with water content ( Figure 7B).

Relationship between Environmental Variables and Biotic Variables
The CCA elucidated the relationship between the environmental variables and biotic variables (Figure 7). Phytoplankton assemblages were distinctively clustered by temperature and salinity, generating a clear segregation between the May and September populations ( Figure 7A). The September phytoplankton communities exhibited a positive correlation with temperature and DIP, while the May communities presented a positive correlation with salinity and DO ( Figure 7A). Particularly, diatoms and dinoflagellates were positively correlated with temperature and DIP ( Figure 7A). In contrast to the phytoplankton assemblages, the CCA results for the dinoflagellate cyst assemblages showed no clear clusters between the two seasons, while COD and AVS were the most influential environmental variables on the dinoflagellate cyst communities ( Figure 7B). The CCA revealed that the gonyaulacoid group was positively correlated with COD and AVS, while the protoperidinioid group was negatively correlated with COD and AVS but positively correlated with water content ( Figure 7B).

Relationship between Water Quality and Phytoplankton Assemblages
Owing to a higher half-saturation constant (Km), diatoms require more nutrients than dinoflagellates, and diatom blooms succeed to dinoflagellate blooms when dissolved inorganic nutrients are limited [61][62][63][64]. After dissolved inorganic nutrients are excessively utilized by diatoms, dissolved organic nutrients (e.g., DON and DOP) become relatively high, and dinoflagellates that can utilize DON and DOP proliferate in coastal waters [65][66][67]. Although silicate was not measured, DIN and DIP levels can infer the growth of diatom populations. During the study period, the phytoplankton assemblages were mainly composed of diatoms and dinoflagellates; particularly, diatoms bloomed in September as a function of a sharp increase in nutrients from May to September. The DIN and DIP levels were significantly different (p < 0.05; Kruskal-Wallis test) between the inner stations located near the aquaculture fish cages and the outer stations connected to the open sea. This is because of the specific characteristics of the study region, where nitrogen and phosphorus compounds from aquaculture farms excessively contribute to the nutrient levels [35].
Multivariate analysis (e.g., CCA) showed the relationship between the environmental variables and phytoplankton assemblages; temperature, salinity, and DO were the most important environmental factors for phytoplankton communities. Consistent with these results, ANN for predicting phytoplankton abundance showed that, in May, the strength of the impact of DO and salinity was relatively large and positive while that of temperature was relatively large and negative (Figure 8). In September, as the water temperature increased above 20 °C, temperature and dissolved inorganic nutrients had a large positive strength of impact on phytoplankton abundance (Figure 8). This is because the increase in water temperature favored phytoplankton growth [68,69], and the decomposition rate of uneaten fish food and fish waste via bacterial activity became vigorous [70].

Relationship between Water Quality and Phytoplankton Assemblages
Owing to a higher half-saturation constant (Km), diatoms require more nutrients than dinoflagellates, and diatom blooms succeed to dinoflagellate blooms when dissolved inorganic nutrients are limited [61][62][63][64]. After dissolved inorganic nutrients are excessively utilized by diatoms, dissolved organic nutrients (e.g., DON and DOP) become relatively high, and dinoflagellates that can utilize DON and DOP proliferate in coastal waters [65][66][67]. Although silicate was not measured, DIN and DIP levels can infer the growth of diatom populations. During the study period, the phytoplankton assemblages were mainly composed of diatoms and dinoflagellates; particularly, diatoms bloomed in September as a function of a sharp increase in nutrients from May to September. The DIN and DIP levels were significantly different (p < 0.05; Kruskal-Wallis test) between the inner stations located near the aquaculture fish cages and the outer stations connected to the open sea. This is because of the specific characteristics of the study region, where nitrogen and phosphorus compounds from aquaculture farms excessively contribute to the nutrient levels [35].
Multivariate analysis (e.g., CCA) showed the relationship between the environmental variables and phytoplankton assemblages; temperature, salinity, and DO were the most important environmental factors for phytoplankton communities. Consistent with these results, ANN for predicting phytoplankton abundance showed that, in May, the strength of the impact of DO and salinity was relatively large and positive while that of temperature was relatively large and negative (Figure 8). In September, as the water temperature increased above 20 • C, temperature and dissolved inorganic nutrients had a large positive strength of impact on phytoplankton abundance (Figure 8). This is because the increase in water temperature favored phytoplankton growth [68,69], and the decomposition rate of uneaten fish food and fish waste via bacterial activity became vigorous [70]. Nutrients originating from aquaculture fish cages were characterized by dissolved and particulate matter. Dissolved nutrients are mostly composed of nitrogen and therefore cause coastal eutrophication because 60% of the released nitrogen is dissolved in the water column. Meanwhile, excessive food and fish waste that are released in the form of particulate matter settle in the surface sediments and are moved to adjacent waters by currents [71]. Only a part of the phosphorus in fish food is assimilated to the fish body and most of the assimilated phosphorus is defecated as fish waste, leading to coastal eutrophication. According to Ackefors and Enell [72], 30% of phosphorous released from uneaten food returned to the fish body by assimilation, and 54% and 16% of the released P compounds are excreted in the form of particulate and dissolved matter, respectively. Eutrophication index (E) was greater than 1 across the sampling stations, an indicator of eutrophication [43,44] and E significantly increased in September as temperature increased (p < 0.05; Kruskal-Wallis test; Table 4). In addition, the DIN and DIP levels in the study region were higher compared to those in the adjacent coastal waters, indicating that the seawater around the Tongyeong fish farms is more eutrophic than other regions along the southern coastal waters of Korea (p < 0.05; Kruskal-Wallis test; Table 5). The DIP levels in the study region were more than twice those in the adjacent coastal waters both in May (p < 0.05) and September (p < 0.01; Kruskal-Wallis test; Table 5); the DIN levels in the study region were significantly higher than those in the other southern coastal waters (p < 0.05; Kruskal-Wallis test; Table 5), indicating that the N and P compounds excreted from the excessive fish food and the fecal and urinary products exacerbated the water quality of the study region. Nutrients originating from aquaculture fish cages were characterized by dissolved and particulate matter. Dissolved nutrients are mostly composed of nitrogen and therefore cause coastal eutrophication because 60% of the released nitrogen is dissolved in the water column. Meanwhile, excessive food and fish waste that are released in the form of particulate matter settle in the surface sediments and are moved to adjacent waters by currents [71]. Only a part of the phosphorus in fish food is assimilated to the fish body and most of the assimilated phosphorus is defecated as fish waste, leading to coastal eutrophication. According to Ackefors and Enell [72], 30% of phosphorous released from uneaten food returned to the fish body by assimilation, and 54% and 16% of the released P compounds are excreted in the form of particulate and dissolved matter, respectively. Eutrophication index (E) was greater than 1 across the sampling stations, an indicator of eutrophication [43,44] and E significantly increased in September as temperature increased (p < 0.05; Kruskal-Wallis test; Table 4). In addition, the DIN and DIP levels in the study region were higher compared to those in the adjacent coastal waters, indicating that the seawater around the Tongyeong fish farms is more eutrophic than other regions along the southern coastal waters of Korea (p < 0.05; Kruskal-Wallis test; Table 5). The DIP levels in the study region were more than twice those in the adjacent coastal waters both in May (p < 0.05) and September (p < 0.01; Kruskal-Wallis test; Table 5); the DIN levels in the study region were significantly higher than those in the other southern coastal waters (p < 0.05; Kruskal-Wallis test; Table 5), indicating that the N and P compounds excreted from the excessive fish food and the fecal and urinary products exacerbated the water quality of the study region.

Relationship between Sediment Environments and Dinoflagellate Cyst Assemblages in Eutrophic Sediments
The extent of eutrophication determines the patterns of organic matter distribution, and the levels of organic matter outstand the southeastern coast of Korea from the Jinju Bay to the Yeongil Bay [73]. The COD levels in the sediments of the fish farms in Tongyeong Sanyang-eup are above the standard level of 20 mg O 2 /g dry [74], indicating that sediment eutrophication in the study region has proceeded due to the massive fish farms and input of sewage and livestock waste from land. The ANN for predicting dinoflagellate cyst abundance also illustrated the strength of the impact of sediment environmental variables on cyst abundance. Largely, COD was the most influential sediment environmental variable while AVS also positively affected cyst abundance in September ( Figure 9). In addition, the ratio of COD to IL (COD/IL) can be utilized to determine the origin of organic matter and the characteristics of sediment distribution, with the organic matter being allochthonous when COD/IL > 1 and autochthonous when COD/IL < 1 [64]. The relatively high COD/IL ratios in the study region (2.83 ± 0.23 in May and 3.19 ± 0.28 in September; Table S3) suggest that the organic matter in the sediments is not likely to have originated from marine aquatic organisms but rather from anthropogenic input such as wastes from fish farms.
In the eutrophic water, heterotrophic/mixotrophic cysts have been detected more than autotrophic cysts, with a high ratio of heterotrophic/mixotrophic species to autotrophic species [16,19,75,76]. In accordance with this, the abundance of heterotrophic/mixotrophic cysts was higher than that of autotrophic cysts across all stations in our study, constituting 62% and 58% of total dinoflagellate cysts in May and September, respectively, inferring that eutrophication has been worsened due to the input of organic matter from the fish farms. Previous studies revealed that the number of species and abundance of dinoflagellate cysts in the eutrophic sediments are higher than those in non-eutrophic sediments [20,77], which is consistent with previous study results from coastal sediments in Korea [14,15,[78][79][80]. While 18 genera and 32 species were identified in this study, previous studies have shown that the number of dinoflagellate cysts in the southern coastal sediments of Korea includes In the eutrophic water, heterotrophic/mixotrophic cysts have been detected more than autotrophic cysts, with a high ratio of heterotrophic/mixotrophic species to autotrophic species [16,19,75,76]. In accordance with this, the abundance of heterotrophic/mixotrophic cysts was higher than that of autotrophic cysts across all stations in our study, constituting 62% and 58% of total dinoflagellate cysts in May and September, respectively, inferring that eutrophication has been worsened due to the input of organic matter from the fish farms. Previous studies revealed that the number of species and abundance of dinoflagellate cysts in the eutrophic sediments are higher than those in non-eutrophic sediments [20,77], which is consistent with previous study results from coastal sediments in Korea [14,15,[78][79][80]. While 18 genera and 32 species were identified in this study, previous studies have shown that the number of dinoflagellate cysts in the southern coastal sediments of Korea includes 2 genera and 27 species in the Jinhae Bay [78], 17 genera, and 30 species in the Gwangyang Bay [15], and 19 genera and 30 species in Geoje [81].
Generally, total cyst abundance in the study region was relatively high and the abundance sharply increased at the outer stations (4820 ± 100 cysts/g at the inner stations vs. 6720 ± 220 cysts/g at the outer stations). As previously mentioned, currents move organic matter originating from aquaculture fish farms to adjacent waters [71]. The fish farms in Tongyeong are employed in coastal waters with fast currents due to the efficient removal of the excessive organic matter [33]. As currents relocated organic matter, dinoflagellate cysts in the surface sediments were also transferred in the direction of the current movement; therefore, cysts accumulated at the outer stations, resulting in higher cyst abundance.
Interestingly, the genus Alexandrium was identified in the study region in September, including Alexandrium affine and Alexandrium catenella/pacificum (Alexandrium pacificum = formerly Alexandrium tamarense). The detection level of paralytic shellfish poisoning (PSP) toxins caused by Alexandrium blooms has already exceeded the federal closure limit in this region, and closure of the harvest bay annually occurs in May [82]. Recently, the extent of the toxin level has aggravated [83], and the PSP toxins also caused human deaths in the 1980s and 1990s in Geoje and Busan, Korea [82,84]. The occurrence of favorable conditions for cyst germination enables red tides, which drive fish mortality in aquaculture farms and further threaten the health of human beings [85]. Given that a moderate abundance Generally, total cyst abundance in the study region was relatively high and the abundance sharply increased at the outer stations (4820 ± 100 cysts/g at the inner stations vs. 6720 ± 220 cysts/g at the outer stations). As previously mentioned, currents move organic matter originating from aquaculture fish farms to adjacent waters [71]. The fish farms in Tongyeong are employed in coastal waters with fast currents due to the efficient removal of the excessive organic matter [33]. As currents relocated organic matter, dinoflagellate cysts in the surface sediments were also transferred in the direction of the current movement; therefore, cysts accumulated at the outer stations, resulting in higher cyst abundance.
Interestingly, the genus Alexandrium was identified in the study region in September, including Alexandrium affine and Alexandrium catenella/pacificum (Alexandrium pacificum = formerly Alexandrium tamarense). The detection level of paralytic shellfish poisoning (PSP) toxins caused by Alexandrium blooms has already exceeded the federal closure limit in this region, and closure of the harvest bay annually occurs in May [82]. Recently, the extent of the toxin level has aggravated [83], and the PSP toxins also caused human deaths in the 1980s and 1990s in Geoje and Busan, Korea [82,84]. The occurrence of favorable conditions for cyst germination enables red tides, which drive fish mortality in aquaculture farms and further threaten the health of human beings [85]. Given that a moderate abundance of Alexandrium cysts appeared in the study region and cysts can play a role as seed populations for blooms [86], continuous monitoring is necessary to detect red tides caused by Alexandrium species.

Relationship between Phytoplankton and Dinoflagellate Cysts
Heterotrophic/mixotrophic dinoflagellates utilize diatoms as a food source for growth [87][88][89]. Protoperidinium species of heterotrophic dinoflagellates feed on a variety of diatoms during diatom blooms or take up excreted dissolved/decaying organic matter from diatoms after the blooms [90,91]. Gaines and Taylor [87] and Jacobson and Anderson [88] described a feeding mechanism in which heterotrophic dinoflagellates deploy pseudopods to completely surround relatively large diatoms and then dissolve their cell contents. In concert with the feeding behavior of heterotrophic dinoflagellates, the seasonal abundance of heterotrophic/mixotrophic dinoflagellate cysts is positively proportional to diatom abun-dance, which was determined using a sediment trap in a prior study [92]. For this reason, heterotrophic/mixotrophic dinoflagellate cysts are associated with the abundance of diatoms, while heterotrophic/mixotrophic cysts dominate in highly productive regions such as upwelling regions [93][94][95]. Consistent with this, our study showed a positive correlation of dinoflagellate cysts with phytoplankton abundance (R 2 = 0.33; p < 0.05; linear regression; Figure 10A), and consequently, the abundance of heterotrophic/mixotrophic dinoflagellate cysts significantly increased in September in accordance with a significant increase in diatom abundance ( Figure 10B). This is because the intrusion of excessive nutrients from aquaculture farms and higher temperature led to the fast growth of diatoms in the warm season (September) and then drove the formation of more heterotrophic/mixotrophic dinoflagellate cysts in the modern sediments. The vertical profile of the dinoflagellate cysts in the study region was not investigated; however, the long history of aquaculture farms in the study region (>30 years) suggests that a long-term process of eutrophication might also have resulted in a relatively high abundance of heterotrophic/mixotrophic dinoflagellates in the past, and a general link between diatoms and heterotrophic/mixotrophic dinoflagellate cysts in relatively recent sediments in the study region.
growth [87][88][89]. Protoperidinium species of heterotrophic dinoflagellates feed on a variety of diatoms during diatom blooms or take up excreted dissolved/decaying organic matter from diatoms after the blooms [90,91]. Gaines and Taylor [87] and Jacobson and Anderson [88] described a feeding mechanism in which heterotrophic dinoflagellates deploy pseudopods to completely surround relatively large diatoms and then dissolve their cell contents. In concert with the feeding behavior of heterotrophic dinoflagellates, the seasonal abundance of heterotrophic/mixotrophic dinoflagellate cysts is positively proportional to diatom abundance, which was determined using a sediment trap in a prior study [92]. For this reason, heterotrophic/mixotrophic dinoflagellate cysts are associated with the abundance of diatoms, while heterotrophic/mixotrophic cysts dominate in highly productive regions such as upwelling regions [93][94][95]. Consistent with this, our study showed a positive correlation of dinoflagellate cysts with phytoplankton abundance (R 2 = 0.33; p < 0.05; linear regression; Figure 10A), and consequently, the abundance of heterotrophic/mixotrophic dinoflagellate cysts significantly increased in September in accordance with a significant increase in diatom abundance ( Figure 10B). This is because the intrusion of excessive nutrients from aquaculture farms and higher temperature led to the fast growth of diatoms in the warm season (September) and then drove the formation of more heterotrophic/mixotrophic dinoflagellate cysts in the modern sediments. The vertical profile of the dinoflagellate cysts in the study region was not investigated; however, the long history of aquaculture farms in the study region (>30 years) suggests that a long-term process of eutrophication might also have resulted in a relatively high abundance of heterotrophic/mixotrophic dinoflagellates in the past, and a general link between diatoms and heterotrophic/mixotrophic dinoflagellate cysts in relatively recent sediments in the study region. Figure 10. Relationship between phytoplankton abundance and dinoflagellate cyst abundance. (A) Linear regression of phytoplankton abundance and dinoflagellate cyst abundance in the study region in May and September. Red indicates abundance in May, and blue indicates abundance in September. The abundance of phytoplankton and cysts in September was significantly higher than that in May (Wilcoxon rank-sum test; p < 0.05). As phytoplankton abundance increased from September to May, dinoflagellate cyst abundance also increased proportionally. (B) Comparison of diatom and heterotrophic/mixotrophic cyst abundance in May and September. A Wilcoxon rank-sum test was performed to compare the difference in abundance between May and September (p < 0.001 for diatoms, p < 0.5 for heterotrophic/mixotrophic dinoflagellate cysts). Hetorocyst = heterotrophic/mixotrophic dinoflagellate cysts. Figure 10. Relationship between phytoplankton abundance and dinoflagellate cyst abundance. (A) Linear regression of phytoplankton abundance and dinoflagellate cyst abundance in the study region in May and September. Red indicates abundance in May, and blue indicates abundance in September. The abundance of phytoplankton and cysts in September was significantly higher than that in May (Wilcoxon rank-sum test; p < 0.05). As phytoplankton abundance increased from September to May, dinoflagellate cyst abundance also increased proportionally. (B) Comparison of diatom and heterotrophic/mixotrophic cyst abundance in May and September. A Wilcoxon rank-sum test was performed to compare the difference in abundance between May and September (p < 0.001 for diatoms, p < 0.5 for heterotrophic/mixotrophic dinoflagellate cysts). Hetorocyst = heterotrophic/mixotrophic dinoflagellate cysts.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/jmse9040362/s1, Table S1: Statistical information of canonical correspondence analysis on phytoplankton assemblages, Table S2: Statistical information of canonical correspondence analysis on dinoflagellate cyst assemblages, Figure S1: Scree plot from canonical correspondence analysis on phytoplankton assemblages, Figure S2: Scree plot from canonical correspondence analysis on dinoflagellate cyst assemblages, Table S3