**3. Results**

### *3.1. Comparative Estimation of Coastal Biota between eDNA Metabarcoding and CMI*

A variety of plankton communities were observed through the two identification methods in Gwangyang Bay. There were differences in the number of identified communities between the two methods (Table 1). In terms of quantity, eDNA metabarcoding seemed to be capable of detecting more species. The average numbers of observed (identified and unidentified) species from the eDNA samples were 27.9 (min to max: 20–36) in June and 49.8 (min to max: 13–72) in September, while those from the CMI were 19.6 (min to max: 12–23) and 18.9 (min to max: 12–24), in June and September, respectively (Table 1). Albeit comparing only with the identified species, we found that the number of species was higher in the eDNA samples than the CMI. On the other hand, in terms of the capability of identification of the eDNA metabarcoding, the unidentified species groups comprised 38% in June and 19% in September (Table 1). Accordingly, in Gwangyang Bay, the eDNA samples identified more species in a higher proportion in September.

In the eDNA samples, the richness values in September were as twice high as those in June (Table 1). The number of identified species was lower in June (mean ± S.D.: 20.2 ± 3.3) than in September (41.7 ± 15.6), which was quite consistent across the study sites. In addition, this pattern was similarly observed from diversity values of the eDNA samples (averages: 1.0 in June and 2.0 in September). The spatial variation of the richness was also lower in June (coefficient of variation: ca. 15%) than in September (ca. 40%). Namely, the heterogeneity of plankton distribution became large in late summer. On the contrary, in the CMI samples, the species richness did not differ between early and late summer; the values of mean and S.D. were 19.6 ± 3.1 in June, and 18.9 ± 3.5 in September. Notably, the level of diversity was comparatively higher in June (2.3 ± 0.2) than in September (1.6 ± 0.2), which was counter to the diversity pattern from the eDNA samples. Considered as a whole, the temporal changes

of biological communities seem to be more distinct, compared to their spatial variation. Nonetheless, we also note some discrepancy of the results in diversity between the two identification methods.


**Table 1.** Richness and Shannon diversity of the samples between water eDNA and conventional microscopic identification (CMI) in Gwangyang Bay. The numbers in the brackets indicate the number of unidentified groups.

To evaluate the consistency of detection and identification of marine plankton groups, we compared the di fferences of community composition between eDNA and CMI samples (Figure 3a,b). Although various groups were detected by eDNA metabarcoding, the community composition was based on the identified groups in eDNA samples, in comparison with those from the CMI samples. In the higher rank of taxonomical classification (>phylum), the eDNA samples comprised 28% of phytoplankton (i.e., algae) and 15% of zooplankton (i.e., Copepoda) (Figure 3a), whereas the CMI samples showed 64% of zooplankton (Figure 3b).

In the eDNA samples, the dominant groups in phytoplankton were diatoms (e.g., *Thalassiosira* spp.) and dinoflagellates (*Hematodinium* spp.). In zooplankton, the dominant groups were marine calanoid copepods such as *Acartia* spp. and *Centropages* spp. in Gwangyang Bay. Crustacea occupied 17% of the identified species, and were primarily comprised of Amphipoda (e.g., *Caprella* spp.), Cirripedia (e.g., barnacles), and Decapoda (e.g., *Corophium* spp.). Cnidaria and Mollusca also engaged species richness of 24% in our study area (Figure 3a). The former consisted mainly of small polyp stony coral, such as *Acropora* spp., and the latter mostly comprised bivalves, such as *Crassostrea* spp. and *Musculista* spp. In addition, several groups, which were relatively low proportionally in CMI, were also well identified, including Annelida (5%), Chaetognatha (4%), Echinoderma (3%), and fish (4%). Particularly for fish, the identification of fish species was quite limited in the eDNA samples, and hence only three genera were identified (*Arnoglossus*, *Engraulis*, and *Oryzias* spp.).

By comparison, the CMI samples showed di fferent proportion in species richness (Figure 3b). The main composition (64%) of zooplankton comprised Cladocera (e.g., *Evadne* spp. and *Podon leuckarti*) as well as Copepoda (e.g., 15 calanoid species and three cyclopoid species). Conversely, a limited number of phytoplankton was identified in the CMI samples, compared to the eDNA samples. The identified phytoplankton were mostly dinoflagellates which were mainly *Noctiluca scintillans*. Crustecea occupied 11% of species richness. Similar to the identified species from the eDNA samples, they were primarily composed of Amphipoda, Cirripedia, and Decapoda. However, most of them were in forms of larvae which was unable to be identified specifically in the CMI samples. Other specific groups were observed

in a small proportion (3%: Annelida, Chaetognatha, Cnidaria, and Echinoderma, and 5%: Fish and Mollusca, see Figure 3b). Nevertheless, in a finer resolution, there was some commonality of species groups between eDNA and CMI samples (Table 2). In both samples, several genera including *Acartia*, *Acropora* and *Centropages*, were commonly observed. At the Gwangyang Bay, *Acartia* spp. were commonly predominant in early summer, while *Centropages* spp. were relatively predominant in late summer. *Acropora* spp. were primarily observed from the eDNA samples around the inner bay in early summer. At the outer bay, including at site 14 and site 15, a dinoflagellate group of *Hematodinium* was relatively abundant, especially in the eDNA samples. In contrast, *Hematodinium* was not detected by CMI in the same area. Moreover, *Oithona* spp. were most predominant in this area, but were relatively less abundant in the outer bay, compared to the inner bay.

**Figure 3.** Community composition from the samples between (**a)** eDNA metabarcoding and (**b**) conventional microscopy identification (CMI). The scatter plots indicate the relationships between species richness and Shannon–Weaver diversity on (**c**) eDNA, (**d**) CMI, and those (**e**) between richness and (**f**) between diversity of the two methods, respectively.


**Table 2.** Dominant plankton groups observed during the summer season (June and September) in Gwangyang Bay.

### *3.2. Relationships of Biotic Information between eDNA and CMI Samples*

To examine consistency of biological information between different sampling strategies, the relationships between species richness and diversity were comparatively assessed. In both eDNA and CMI samples, species richness and diversity were positively correlated with each other (Figure 3c,d). The eDNA samples showed stronger signal of the positive relationship between species richness and diversity than the CMI samples, and the interpretability of species richness on corresponding diversity was three times higher in the eDNA samples (*r*2 = 0.74) than in the CMI samples (*r*2 = 0.23). Although both samples showed the significant relationships between the two, the relationship was clearer in the eDNA samples. On the other hand, we also examined the relationships between the richness values and between the diversity values (Figure 3e,f). There was no statistical significance between the richness values (i.e., eDNA versus CMI samples) (Figure 3e). In addition, although the diversity values exhibited statistical significance in their relationship, the signal was slightly negative, which was counterintuitive (Figure 3f). In consequence, it appeared that the information obtained from the same methodology was consistent enough to project the relationship between species richness and diversity. Conversely, it was found that there was a discrepancy of biotic information between eDNA and CMI samples.

### *3.3. Assessment of Biogeochemical Characteristics in Gwangyang Bay*

The clustering analysis using the SOM characterized biogeochemical features of Gwangyang Bay into four distinct patterns. The four clusters determined by the SOM shaped spatiotemporal heterogeneity of the data samples at Gwangyang Bay (Figure 4 and Appendix B). It is remarkable to discern the spatiotemporal pattern that cluster 1 included site 1 to site 6 of June, cluster 2 site 7 to site 15 of June, cluster 3 site 1 to site 8 of September, and cluster 4 site 9 to site 15 of September as well as site 14 and site 15 of June (Figure 4a). In addition, the estimate of neighboring distances among the clusters indicated that the clusters were firstly separated as top (cluster 3 and cluster 4) and bottom (cluster 1 and cluster 2). As a consequence, the clustering result manifested that plankton community of Gwangyang Bay was primarily characterized by seasonal influences between early and late summer (i.e., June and September at Gwangyang Bay), and then was spatially distinguished. Strictly speaking, site 14 and site 15 of June were grouped as cluster 4 which represented the outer bay of late summer, but they were placed on the bottom of cluster 4, which was characterized as the outer bay of early summer. Namely, these two sites appear to represent similar features on coastal plankton community, regardless of temporal changes in summer.

**Figure 4.** Clustering result (**a**) of the data of water eDNA and CMI based on the self-organizing map. The right panels (**b**) present the corresponding physical, chemical, and biological conditions in Gwangyang Bay. The horizontal lines of zero indicate corresponding grand average values (water temperature: 25.6 °C, salinity: 29.3 psu, TP: 0.049 mg <sup>L</sup>−1, TN: 0.45 mg <sup>L</sup>−1, TC: 22.4 mg <sup>L</sup>−1, Chl-*a*: 4.36 mg <sup>L</sup>−1).

Several water quality parameters delineated ambient physicochemical conditions associated with plankton community in Gwangyang Bay (Figure 4b). Water temperature was relatively lower in cluster 2 and higher in cluster 3 among the four groups. The higher salinity of the outer bay matched well with cluster 2 and cluster 4. Cluster 3 represented the inner bay of the summer, exhibiting lower salinity was higher water temperature. Concerning nutrient concentration, total phosphorus (TP) concentrations were higher in June (cluster 1 and cluster 2 in Figure 4) than in September (cluster 3 and cluster 4 in Figure 4). In the spatial scale, TP was higher at the inner bay (cluster 1 and cluster 3 in Figure 3) than at the outer bay (cluster 2 and cluster 4 in Figure 4). In addition to TP, total nitrogen (TN) concentrations were conspicuously high in cluster 3, which represented the inner bay in late summer. Total carbon (TC) concentrations displayed opposite patterns against TN. Among the four clustering groups, chlorophyll *a* (Chl-*a*) concentrations were highest at the inner bay in early summer, while were lowest at the outer bay in late summer. In view of biotic information, the number of species was relatively higher in cluster 3 and cluster 4 (September) based on the eDNA samples, while the diversity indices were comparatively higher in cluster 1 and cluster 2 (June) based on the CMI samples (Figure 4).
