Assessment of fish diversity in the coastal waters off Nodaedo Island, Tongyeong, Korea, using an underwater visual census and environmental DNA metabarcoding

Environmental DNA (eDNA) analysis is a promising method for monitoring marine biodiversity. However, it is necessary to verify whether eDNA analysis provides an accurate measure of biodiversity in species-rich ecosystems such as coastal waters as part of a long-term monitoring campaign. Therefore, both an underwater visual census (UVC) and eDNA metabarcoding were conducted monthly from June 2018 to May 2019 to investigate fish diversity off the coast of Nodaedo Island, Tongyeong, South Korea. A total of 16,036 fish were observed in the UVC and classified into 69 species. Perciformes was the most dominant group with 31 species. eDNA metabarcoding identified 68 operational taxonomic units and 738,217 reads, encompassing 18 orders, 42 families, and 68 species. The Gobiidae family was the most dominant taxon with five species. The UVC was able to detect species that were otherwise not identified using eDNA due to the low number of individuals or analytical limitations, while eDNA metabarcoding was suitable for detecting smaller species, pelagic species, and those that inhabit microhabitats. Our findings demonstrated that these two methods can be used to complement each other and gain insights into the structure of fish communities in various coastal environments. Though eDNA metabarcoding represents a fast and efficient method for surveying fish species diversity in coastal waters, it should be carefully compared with other traditional survey methods to ensure that it is suitable as an effective complementary measure.


Introduction
With more than 30,000 marine fish species worldwide, teleosts represent the largest and most diverse group of vertebrates; thus, fish diversity is often used as an indicator to monitor marine ecosystems (Thomsen et al. 2012;Froese and Pauly 2022).Measurements of the diversity of marine fish can be used to monitor climate change, protect fishery resources, and support environmental conservation management (Roessig et al. 2004;Mora et al. 2008).This is particularly important for coastal areas, which are an ideal breeding ground for adult fish and a refuge for larval and juvenile fish and other organisms such as mollusks and crustaceans (Parrish 1989;Mateo and Tobías 2001) due to the extensive rock and seaweed cover and high levels of organic matter.For example, the southeastern waters of South Korea are influenced by the Tsushima Warm Current, which is a tributary of the Kuroshio Current that is characterized by high temperatures and high salinity.This region exhibits high fish productivity due to the continental shelf and the development of reclaimed coastline (KORDI 1998;Guo et al. 2006).Nodaedo is an island off the coast of Tongyeong in southeastern Korea that is located in an area with higher water temperatures and salinity than observed for the rest of the coastal area; thus, it also has high fish diversity (Choo and Kim 1998;Kim et al. 2016).
Most current methods employed to survey fish diversity (e.g., tow nets and beam trawling) are time-consuming, destructive, and can only be used in specific areas.In addition, these approaches are not well suited to evaluate fish diversity along shallow coasts.As such, underwater visual censuses (UVCs) is a well-known method for the assessment of fish diversity in shallow coastal waters.Unlike conventional approaches that rely on the use of fishing equipment, this method is not only fast, non-disruptive, and highly effective, but can also be employed in various environments (Thresher and Gunn 1986;St. John et al. 1990;Lowry et al. 2012).However, due to variations in diving technology and fish observability caused by water turbidity, it can be difficult to observe certain fish species, such as those with a narrow habitat or those primarily found near the surface (Watson et al. 1995;Thompson and Mapstone 1997;Kulbicki 1998).
Environmental DNA (eDNA) is DNA derived from the hair, cells, feces, and tissues of organisms that enter surrounding environmental media, such as the soil, air, and water (Taberlet et al. 2012;Mächler et al. 2016).Ficetola et al. (2008) were the first to apply eDNA technology to aquatic environments for monitoring bullfrog (Rana catesbeiana) in ponds, and various studies have since employed eDNA to survey aquatic organisms (Shaw et al. 2016;Grey et al. 2018;Kelly et al. 2018).Of the eDNA analysis methods currently available, eDNA metabarcoding is particularly useful because it allows for the rapid analysis of large numbers of eDNA sequences to identify the species present at a specific location, leading to highly accurate biological classifications (Yu et al. 2012;Wood et al. 2013).Recently, the use of both eDNA analysis and underwater survey methods (e.g., UVCs, snorkeling, and baited underwater video censuses) has become common in fish species diversity research, and there appears to be a complementary relationship between these two methods (Ulibarri et al. 2017;Stat et al. 2019;Valdivia-Carrillo et al. 2019).For example, eDNA metabarcoding is suitable for detecting migratory and active fish species that are difficult to observe using a UVC, whereas a UVC is suitable for observing species that cannot be detected using eDNA analysis because of the low number of individuals.In addition, a UVC allows the size, sex, behavioral characteristics, and number of individual fish to be recorded, which cannot be determined using eDNA.On the other hand, eDNA analysis ensures accurate species identification through molecular analysis and can cover a wider area in a shorter time by collecting water samples.Therefore, more fish can be detected when these two methods are combined than if they are used alone.However, to the best of our knowledge, no previous studies have characterized the diversity of fish species in Korean coastal waters using both UVC and eDNA metabarcoding.Therefore, our study aimed to employ these two methods to explore the fish diversity around Nodaedo Island, Tongyeong, South Korea.

Study area and sample collection
The survey was conducted in the coastal waters off Nodaedo Island, approximately 25 km from Tongyeong, South Korea (Fig. 1).UVC surveys and eDNA water sampling were conducted at the same site once a month from June 2018 to May 2019.These surveys were conducted 2 days after neap tide to ensure visibility during the UVC.

Underwater visual census
The UVC was conducted using the line transect method to observe species within a certain range while scuba diving in groups of two.The line transect created an observation area with a length of approximately 200 m and an area of 1,200 m 2 .Underwater photos and videos were taken using a DSC-RX100 camera (Sony, Japan) and a GoPro Hero 4 Black camera (GoPro, USA) for approximately 40 min during each UVC survey.The species were identified in accordance with Kim et al. (2005), with the images and photos compared at the species level in the laboratory.FishBase (Froese and Pauly 2022) was used to determine the classification system, scientific names, and climate categories.The number of fish in the photos and videos was counted using the Image J program (NIH, USA).The water temperature was measured using a ZOOP Novo dive computer (Suunto, Finland).Salinity data (psu) were obtained from the Korean Hydrographic and Oceanographic Agency (KHOA).Each month, five seawater samples (200 mL per sample) were collected for eDNA analysis approximately 1.5 m above the sea bottom using a stand-up plastic pouch at 50-m intervals along a 200-m transect line.The five samples were combined (1 L) and placed in a container treated with 10% commercial bleach and passed through a Sterivex-HV filter (0.45 µm) using a 250-mL syringe.After filtering the seawater, the Sterivex-HV filter was dried using a 250-mL syringe.The Sterivex-HV filters were then capped and stored in a cooler with dry ice and transported to the laboratory within an hour prior to DNA extraction.We also processed three field blanks consisting of 1 L of distilled water as a negative control using the same method.To minimize contamination, latex gloves were worn during filtration and all tools were disinfected with chlorine.
Neither the negative nor extraction control demonstrated any evidence of amplification.
To quantitatively assess the amplicon library, the concentration was determined using the Synergy H1 and Quan-tiFlour dsDNA systems (Promega, Madison, WI, USA; Table S1).The library quality was assessed using a fragment analyzer (Advanced Analytical Technologies, Ankeny, IA, USA) with a dsDNA 915 Reagent Kit (Agilent, Santa Clara, CA, USA) following the manufacturer's instructions.Paired-end sequencing (2 × 300 bp) was conducted on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) with a MiSeq Reagent Kit v3.

eDNA and UVC data analysis
Reads beginning with a sequence that completely matched the primer were extracted using the fastx_barcode_splitter tool within FASTX-Toolkit (v.0.0.14), after which the primer sequence was trimmed.The primer array and 70 bases in the second half of the read were removed from the extracted array.Afterward, sequences with quality scores below 20 were removed using the Sickle tool (Joshi and Fass 2011), while sequences with a length of ≥ 40 bp and fewer than 100 assigned reads were discarded.The reads were merged using 180 bases with the FLASH (v.1.2.11) paired-end merge script (Magoč and Salzberg 2011).The sequences were then clustered into operational taxonomic units (OTUs) using USEARCH (Version 7.0.1090)(Rognes et al. 2016) with a 97% similarity threshold.Chimeras were filtered out using UCHIME (v4.2.40) (Rognes et al. 2016).Systematic estimation was then conducted by comparing Mitochondrial Genome Database of Fish (MitoFish) and MiFish references using the BLAST algorithm.The read and eDNA metabarcoding data were converted into presence/absence data to enable a comparison with the UVC data.Bray-Curtis similarity analysis was conducted to assess the similarities between the fish communities identified using the two survey methods.One-way analysis of similarity (ANOSIM) was used to analyze the differences in the fish communities between the survey methods.Differences in method and season were tested by performing PERMANOVA analysis using the function adonis (package vegan) and pairwiseAdonis (Martinez Arbizu 2020).Ordination by non-metric multidimensional scaling (NMDS) based on Bray-Curtis similarity was used to visualize the dissimilarity in the community structure, and similarity percentage (SIMPER) analysis was conducted to find the species responsible for these differences.All of these analyses were performed using the PRIMER v6 statistical package (Plymouth Routines in Multivariate Ecological Research, v6.1.6;PRIMER-E Ltd., Lutton, Ivybridge, UK) and the "vegan" Page 4 of 15 package (Oksanen 2013) in R version 4.1.2(R Core Team 2013), while the graphs were constructed using ggplot2 in R.

Temperature and salinity
The highest water temperature was 25 °C recorded in August 2018 and the lowest was 11 °C observed in January 2019.The average salinity was 33 psu (Fig. 2).

Fish species detected using eDNA metabarcoding
A total of 738,217 reads were analyzed using eDNA metabarcoding, resulting in the identification of 18 orders, 42 families, and 68 species (Table S2).Of these, Gobiidae appeared five times, Clupeidae and Labridae each appearing four times, followed by Scombridae, Carangidae, Monacanthidae, Stichaeidae, Engraulidae, and Pleuronectidae which appeared three times.One or two species were identified for each of the remaining 32 families.Some sequences had very high similarities, including Sebastes spp., Hexagrammos spp., Ditrema spp., Takifugu spp., Scomber spp., Platycephalus spp., Gadus spp., Thunnus spp., and Pseudopleuronectes spp.Therefore, they were not classified at the species level in the OTU analysis (Table S2).
Of the taxa classified at the species level, Engraulis japonicus was identified in nine of the twelve sampling months, while Pagrus major and Trachurus japonicus were recorded eight times, Ammodytes personatus, Clupea pallasii, and N. ransonnetii seven times, and P. poecilepterus, Lateolabrax japonicus, P. rubripinnis, and Stephanolepis cirrhifer six times.The month with the highest number of species was January 2019 (26 species), when the water temperature was the lowest, followed by 25 species in November 2018.In contrast, only five species were identified in April 2019 and nine in February.These results are in contrast to those obtained from the UVC.

Comparison with previous research
Overall, the present study found a higher species diversity than recorded by previous studies based on UVCs (Table 1); of the 105 species detected across both sampling methods, 69 species were identified using the UVC and 68 species using eDNA metabarcoding.Two previous studies using the same method (UVC) reported the presence of 59 species (Gwak et al. 2016;Lee et al. 2018), while our study shows higher numbers at both the species and individual levels.Twenty-two species, including D. temminckii, N. ransonnetii, Hexagrammos spp., Takifugu spp., P. zonoleucus, and R. ercodes were commonly detected by UVCs and eDNA metabarcoding in the present study and the previous studies (Fig. 3).These species are known to inhabit coastal habitats and are mainly characterized as sedentary or school-forming species.On the other hand, the species observed only in previous studies were 13 in total, such as Hippocampus spp., Chelidonichthys spinosus, Urocampus nanus, Upeneus japonicus.These were mostly species with a low frequency of occurrence and a small number of individuals.The species observed only in eDNA and UVC were 31 and 17, respectively.In eDNA surveys, we found fast-swimming Scomber spp.and Thunnus spp., which are difficult to observe with UVC methods.In contrast, the majority of species, including Oplegnathus punctatus, P. coelestis, and V. centropomus observed only through UVC were similar to previous studies.They also showed a low frequency of occurrence and a small number of individuals.
In the present study, fish species that had not previously been recorded off the coast of Korea in the South Sea were also observed (Table 2).For example, S. bapturus, a fish species mainly observed around Ulleungdo and Dokdo Island in the East Sea, was recorded for the first time in Tongyeong attached to the bedrock.Parupeneus indicus, which is classified as an endangered species on the IUCN red list (Smith-Vaniz and Williams 2016) was also observed.

Results from the UVC and eDNA metabarcoding
ANOSIM analysis identified a significant difference between the eDNA metabarcoding and UVC data (ANOSIM, global R = 0.695, P = 0.001).Based on the species contribution (%), SIMPER analysis revealed that the average dissimilarity between eDNA metabarcoding and the UVC was 75.94% (Table 3).Scomber spp.and E. japonicus were not observed in the UVC, while C. notata, P. zonoleucus, and P. sieboldi were detected in at least nine of the monthly UVC surveys but only once or twice in the monthly eDNA metabarcoding samples.In addition, nMDS analysis of both methods using monthly presence/absence data produced two groups that were distinguished by the sampling method.However, according to the PERMANOVA analysis, seasonal differences were not observed beyond spring versus fall (R 2 = 0.126, P = 0.032) (Fig. 4).
A total of 22 orders, 54 families, and 105 species were observed using UVC and eDNA metabarcoding combined (Table 4).A total of 32 species were observed with both  methods, 37 species were observed only in the UVC, and 36 species were detected only using eDNA metabarcoding.The largest difference between the monthly number of species observed using the two methods was found in September and November 2018.Examining the monthly appearance patterns of six species (Hexagrammos spp., N. ransonnetii, P. rubripinnis, P. poecilepterus, Sebastes spp., S. cirrhifer) that were detected in both survey methods and appeared more than five times, Hexagrammos spp.and P. rubripinnis showed dissimilar patterns between the two methods.However, the remaining four species exhibited a similar pattern of appearing in the same month (Fig. 5).
Most of the species identified exclusively using the UVC were slow swimmers that settled on rocks or sand.However, there were some notable exceptions, such as Seriola dumerili, which was observed despite being an agile swimmer, and Ostorhinchus semilineatus, for which a large number of individuals were observed exclusively in the UVC.
Those fish species detected only through eDNA metabarcoding were generally active swimmers with a strong migratory tendency.These fish, which included members of the Clupeidae and Blenniidae families, also tend to shelter in microhabitats that are difficult to observe in a UVC.Species that were observed in both methods generally had strong sedentary characteristics and were less likely to flee in the presence of divers.However, active swimmers such as A. personatus and T. japonicus were exceptions, with both of these species tending to appear together.

UVC results
In the present study, a total of 69 species were observed in the offshore UVC of the coastal waters off Nodaedo Island.
In a previous study of inshore fish in Tongyeong using a UVC, 43 species (Gwak et al. 2016) and 45 species (Lee et al. 2018) were identified annually, much lower than the number of species in our study.In addition, the number of individuals of those species identified in both studies varied noticeably, highlighting the differences in the distribution patterns and structure of the inshore and offshore fish communities.
Of the fish species observed off the coast of Nodaedo Island, Chromis notata was the dominant species, appearing in 12 of the monthly UVC surveys with 7,870 individuals.Huh et al. (2013) (Oh and Noh 2006).Many studies conducted along the south coast of South Korea have reported the presence of a large number of individuals from these species.UVC surveys demonstrated that both S. inermis and C. notata tend to occur in together, which is thought to be because the former is carnivorous and the latter is planktivorous; thus, they do not compete with each other for food (Ochi 1986;Huh and Kwak 1998).
The seasonal occurrence of certain fish species is at least partially related to the water temperature.However, species richness does not necessarily strictly increase with the water temperature because both the present study and previous studies (Gwak et al. 2016;Han et al. 2017;Lee et al. 2018) have recorded a large number of species between September and November, while the highest water temperature is in August.This is likely because the current that travels from Jeju Island to the northeast during summer and autumn transports fish (Kim and Bae 2011).In addition, in the present study, the number of species observed in August was lower than that observed in July and September, primarily due to the lack of visibility during the diving survey.UVCs are a nondestructive method that is widely used in rocky areas, but underwater visibility can severely affect the accuracy of this approach, which in turn depends on the weather and/or environmental conditions.Gwak et al. (2016) reported water temperatures 7 ℃ in January and February and 29 ℃ in July and August.In contrast, the water temperature range in this study was 11 ℃ in January and 25 ℃ in August.Therefore, water temperature does not fluctuate as dramatically around Nodaedo Island as in the Bay of Tongyeong (Kim et al. 2016).The eastern waters of the South Sea are affected by a mix of Tsushima Warm Current waters, cold water from the East Sea, and coastal waters of the South Sea, resulting in complex marine characteristics.In particular, a water temperature front has formed between the waters around Nodaedo Island and the bay of Tongyeong.Therefore, the area surveyed in the present study represents a more stable environment than the bay of Tongyeong.
Most of the species observed only in the UVC were represented by only a small number of individuals with a low occurrence frequency.Takahara et al. (2012) reported a significant relationship between biomass and the eDNA concentration for living organisms.Therefore, the number of individuals for a given species can be so low that their eDNA becomes undetectable.However, although the number of individuals of Ostorhinchus semilineatus increased considerably from September to November compared to other species, this increase was only detected using the UVC.This was likely because the MiFish primers may not be the optimal choice for the amplification of the 12S gene for this species.
Most of the subtropical and tropical fish species that were observed off the coast of Nodaedo Island appeared between September and December, with only one or two individuals per species.Lee et al. (2018) also reported mostly subtropical fish species during the same period.In general, the appearance of subtropical and tropical fish species on the temperate coast is likely to be a result of transport from other locations due to ocean currents (Nakamura et al. 2012).At the end of summer, the ocean currents in the South Sea begin to travel northeast from Jeju Island (Kim and Bae 2011).Given that the period from June to September is the spawning season for Pomacentrus coelestis, Chaetodontoplus septentrionalis, and Labroides dimidiatus (Sakai and Kohda 2001;Kim et al. 2005;Chen and Tzeng 2009), it is possible that larval and juvenile fish were transported to the coastal waters of Tongyeong.Tropical fish can survive when the water temperature is at least 16-18 ℃ (Figueira and Booth 2010).However, in the study area, the winter water temperature was < 15 ℃, which is unsuitable for the survival of subtropical and tropical fish.Therefore, these fish species were not observed during the winter in the UVC, and these fish were unlikely to have reproduced during this period (Nakamura et al. 2013).In contrast, tropical species may settle on the southern coast if the annual average Fig. 5 A bubble plot for comparing the number of reads obtained through eDNA analysis and the number of individuals for fish species identified using the underwater visual census (UVC) water temperature on the south coast of Korea continues to increase and if other environmental conditions such as the food supply become more favorable for tropical and subtropical fish species (Harriott and Banks 2002;Seong et al. 2010).The subtropical and tropical fish species observed in the UVC surveys were not detected through eDNA metabarcoding.This may be due to the low volume of PCR reactions and the low volume/concentration of DNA added to the PCR reactions.

eDNA metabarcoding
The use of eDNA to identify marine fish species has been a relatively recent with Thomsen et al. (2012) the first to report the use of this method.Compared with traditional methods (e.g., fyke nets, underwater visual censuses, and trawling), the eDNA method has recently been proven to be both time-efficient and cost-effective (Shaw et al. 2016;Thomsen et al. 2016;Hinlo et al. 2017;Yamamoto et al. 2017;Stat et al. 2019;Polanco Fernández et al. 2021).We used both eDNA metabarcoding and a UVC to overcome the limitations of existing methods in assessing fish diversity.The species detected using eDNA metabarcoding were generally those associated with microhabitats, such as Blenniidae and Stichaeidae, which are difficult to observe using a UVC, and those with migratory and schooling behavior, such as C. pallasii, Engraulis japonicus, Ammodytes personatus, and Trachurus japonicus.A complementary relationship with the results from the UVC method was also observed, with eDNA metabarcoding detecting fish species that cannot be easily observed using a UVC.The complementary results of UVC and eDNA methods were similar to those reported in previous studies (Thomsen et al. 2012;Stat et al. 2019;Valdivia-Carrillo et al. 2019).
The eDNA detection results revealed significant seasonal variation in the appearance of different fish species.For example, E. japonicus appeared in nine of the monthly samples and T. japonicus appeared eight times, which is consistent with reports of their continuous presence over all four seasons in the southern seas of Korea (Kim et al. 2011(Kim et al. , 2013)).Furthermore, A. personatus is a cold-water fish species characterized by its migration to the open sea as water temperatures rise during the summer months (Kim et al. 1994).Therefore, it is concluded that A. personatus did not appear in this study during the high-temperature season.Conversely, Parajulis poecilepterus, which is known to inhabit areas influenced by turbulent waters, did not appear during the colder winter months when water temperatures were low (Lee and Kim 1996).In fact, 66% of total eDNA reads for Clupea pallasii were found in January and March, suggesting that its presence may be influenced by the spawning season (Lee et al. 2017).
Overall, the number of species detected each month using eDNA and the water temperature did not appear to be significantly correlated.For example, although the lowest water temperature was in January 2019, this month had the highest number of identified species for the year, with 18 and 8 fish species classified at the family and genus levels, respectively.This trend was found to be significantly differed from the UVC results conducted at the same time.Unlike January 2019, a low number of species was detected in December and February even though there was no significant difference in the water temperature in January, December, and February.This suggests that factors other than water temperature may have influenced the results with the highest number of species recorded in January.In particular, Stat et al. (2019) reported that eDNA detectability largely depends on biological mobility.Therefore, the high eDNA levels of the species detected in January were likely due to the presence of a large number of individuals, high activity, and/or a wide DNA distribution.In fact, most of the highly active migratory fish species such as Sardinops sagax, C. pallasii, Konosirus punctatus, A. personatus, Seriola quinqueradiata, and E. japonicus were present in January.In addition, those fish species that were detected more than six times during the year are generally considered active fish species, such as E. japonicus, T. japonicus, A. personatus, C. pallasii, and Pagrus major, which supports our hypothesis.Overall, our results suggest that eDNA is strongly reflective of the fish species present rather than water temperature, while UVC surveys are believed to provide a snapshot of the winter species composition in limited areas.Scomber spp., which appeared 11 times in the present study, was likely misclassified as Scomber japonicus, which occurs year-round in the South Sea (Lee and Kim 2011).However, S. japonicus has never been observed in a UVC.Davis and Anderson (1989) reported that UVCs may underestimate fish species that swim quickly or that mainly inhabit surface waters.Therefore, mackerel was not observed in the UVC survey in this study.
In this study, the monthly DNA extraction concentration was lower than 0.5 ng/μl in 7 out of 12 samples.Particularly, in March and April, the concentrations were extremely low at 0.140 ng/μl and 0.201 ng/μl respectively (Table S3).Mauvisseau et al. (2019) reported a correlation between the concentration of extracted DNA and the number of species detected.However, our findings that we detected 16 species in March and 5 species in April suggest that it is possible to detect species even with very small amounts of DNA.However, as this study aims to verify biodiversity in a specific area, future analyses should consider factors such as seawater filtration volume and repetition of sampling to detect a greater number of species.
Sebastes spp. was presumed to be S. inermis because it appeared 12 times during the year in the UVC.In the case of 1 3 Hexagrammos spp., it was presumed to be H. otakii and H. agrammus, as it appeared in large numbers in the UVC.In other studies using the same primers employed in the present study, Sebastes spp.and Takifugu spp.have not been classified at the species level (Miya et al. 2015;Aizu et al. 2017;Yamamoto et al. 2017).According to Aizu et al. (2017) and Kelly et al. (2014), the reason some species are not classified at the species level may be due to the primers used or missing mitochondrial DNA data for specific species.Therefore, fish not classified at the species level, the use of various primer sets and/or database validation is necessary.
The fish species detected only via eDNA were either bottom-dwelling like Pholis nebulosa or pelagic fish like C. pallasii.Because the bottom is sandy and muddy at a water depth of > 7 m, P. nebulosa living on the bedrock can be easily detected in the survey area using the eDNA method.However, pelagic and schooling fish are unlikely to inhabit the surveyed waters because they are highly mobile, and some are migratory.Therefore, it is probable that the detected eDNA was transported from outside the study area by tides or currents.In fact, previous research has reported that currents and tides affect the detection results by transporting eDNA outside the survey area (Kelly et al. 2018).Therefore, it is possible that the pelagic fish detected in the present study do not naturally inhabit the survey area, despite their eDNA being detected there (Valdivia-Carrillo et al. 2019).The half-life for the detection of eDNA in an artificial seawater environment has been reported to be 26 h (Collins et al. 2018).However, the decomposition rate of DNA in the sea is expected to be faster because of the effects of pH, water temperature, salinity, UV radiation, and microorganisms (Strickler et al. 2015;Salter 2018).Kelly et al. (2018) compared the eDNA of fish clusters in three regions of Hood Cannal, Washington, USA and reported that seawater flow dynamics (e.g., tidal flow) are responsible for the transport of environmental DNA from non-survey areas, affecting the detection results.Andruszkiewicz et al. (2019) also reported the movement of eDNA at 10 km/day due to ocean currents.This suggests that the pelagic and schooling fish species detected in this study in fact inhabited the southeastern seas of Korea.
In the present study, the ANOSIM analysis results for the presence/absence data revealed significant differences between eDNA metabarcoding and the UVC.The fish species found in both the UVC and eDNA metabarcoding surveys could be classified into three main groups: 1) few individuals and few appearances in the UVC + many eDNA detections; 2) few individuals but many appearances in the UVC + few eDNA detections; and 3) many individuals and many appearances in the UVC + few eDNA detections (Table 5).Acanthopagrus schlegelii and Mugil cephalus were members of the first group, for which it was assumed that there were many more individuals in the study area than those observed in the UVC surveys.The second group included Enneapterygius etheostoma, Rudarius ercodes, Paracentropogon rubripinnis, Semicossyphus reticulatus, Pseudolabrus sieboldi, and Pterogobius zonoleucus, which are believed to have limited detection rates because of their low DNA concentrations.The third group included C. notata, which appeared 12 times in the 12 UVC surveys with a high population number, but it was only detected once in the eDNA metabarcoding samples over the entire survey period.Miya et al. (2015) reported detecting Chromis spp. in a tank using similar procedures as those used in this study.However, the detection of C. notata only once in this study is likely due to insufficient PCR replication or an inadequate volume of filtered seawater.
The UVC method used in the present study is suitable for fish species whose eDNA is not detectable because of a low number of individuals.While eDNA method is suitable for detecting pelagic and highly active fish species.Valdivia-Carrillo et al. (2019) also demonstrated differences in the species composition using the two methods, with eDNA detecting only pelagic and high-activity fish species such as Istiophoridae, Scombridae, Mugilidae, and Clupeidae.In contrast, fish species observed only in the UVC in their study followed a similar pattern to the results of the present study because fish species with strong sedentary behavior were observed.Therefore, it appears that there is a complementary relationship between the two methods.However, to minimize the difference in the species composition between these two methods, the use of high-quality video cameras during UVCs, larger water sample sizes, and/or more PCR replication or a higher sequencing depth during eDNA analysis is recommended.It is difficult to conclude whether the collection of fish species recorded in the present study reflects all of the fish species present off the coast of Nodaedo Island.However, given the seasonal patterns of fish distribution and the transport of eDNA via ocean currents, the results of this study likely reflect not only the biodiversity of Nodaedo Island but also that of nearby areas, such as the southeastern coast of Korea.

Fig. 1
Fig. 1 Map showing the location of survey area in the coastal waters off Nodaedo Island, Tongyeong, Korea.A closed black circle shows the site for UVC and eDNA surveys

Fig. 2
Fig. 2 Monthly variations in water temperature and salinity in the coastal waters off Nodaedo Island, Tongyeong, Korea from June 2018 to May 2019 explained that the appearance of C. notata in the central and southeastern regions of the East Sea was a result of strong currents and climate change.Shin et al. (2014) reported genetic similarities between the C. notata populations around Jeju Island and the East Sea.These observations support the hypothesis that a C. notata population from the coast of Jeju Island may have moved northward due to the influence of water currents.However, C. notata individuals were observed throughout the year, suggesting that C. notata has already adapted to seasonal changes and has settled on the coast of the South Sea.Nevertheless,Lee et al. (2018) investigated the fish diversity in the bay of Tongyeong and reported different results, recording approximately 50 C. notata individuals appearing only 5 times a year.Although the bay of Tongyeong(Lee et al. 2018) and Nodaedo Island are 25 km apart, they have similar water temperatures and salinity.Therefore, the difference in the number of fish and their appearance frequency are likely to be influenced by geographic characteristics

Fig. 4
Fig. 4 Non-metric multidimensional scaling (NMDS) of total fish species per sampling method and season, based on presence/absence transformed data and the Bray-Curtis similarity index

Table 1
Comparison of the present study with previous studies on fish species composition conducted by underwater visual census in the coastal waters off Tongyeong (Gwak et al. 2016;Lee et al. 2018)ber of species data collected from eDNA metabarcoding, UVC, and previous studies(Gwak et al. 2016;Lee et al. 2018)

Table 2
List of tropical and subtropical fish species observed at Nodaedo Island in Tongyeong (F: frequency, M: month)

Table 3
Species contribution (%) to average dissimilarity resulting from SIMPER between eDNA metabarcoding and UVC

Table 4
Classification of fish observed and detected by UVC and eDNA in the coastal waters off Nodaedo Island, Tongyeong, Korea from June 2018 to May 2019 C. notata has possibly adapted more quickly to Nodaedo Island, which is directly affected by the Tsushima Warm Current.These observations suggest that C. notata requires more time to settle in the closed bay of Tongyeong.In contrast, Sebastes inermis, the second-most dominant species, is known to adapt easily to changes in water temperature, and this species has already settled off the coast