eDNA surveys and trace-tracking fieldwork were conducted in Yeouido Saetgang Ecological Park, which is in Seoul, South Korea. Saetgang is a small river that branches off from the Han River and flows around an alluvial island called ‘Yeouido’. The Yeouido Saetgang Ecological Park is a green space that emphasises ecological elements, such as emergent plant communities, eco-friendly ponds and bioswales, with a total length of 4.6 km and a total area of 758,000 m². To identify the typical distribution of the Eurasian otter in Saetgang, both surveys were conducted from the river entrance to the confluence points. Along the river, six sampling sites were located at intervals of approximately 500 m (YS1, YS2, YS4, YS5, YS6 and YS7). Additionally, one sampling site was in a pond connected to the river inside the Ecological Park (YS3) (Fig. 1). To assess the physical environment at each study site, we examined stream width, actual water width and substrate composition. The substrate was assessed in five classes, based on particle size: B (Boulder, > 256 mm), C (Cobble, 64–256 mm), P (Pebble, 4–64 mm), G (Gravel, 2–4 mm) and S (Silt, < 2 mm) (Suppl. material 3).

Figure 1.

The location of seven sampling sites in Yeouido Saetgang Ecological Park.

The eDNA sampling at seven sites in study area was conducted on 27 April (spring) and 11 October 2022 (autumn). To compile an adequate species list for the study area, we used Sterivex cartridge filters with a 0.45 µm pore size (Millipore), assuming replication and filtered 330 ml through each filter for a total of 990 ml of water (Sato et al. 2017; Dickie et al. 2018). In addition, we filtered the water in the vivarium of Eurasian otters at Korean otter Research Center as the positive eDNA sample. Contamination occurring during the sampling process was monitored using negative controls. Independent sampling materials including cartridge filter and 30 ml syringe were used to prevent contamination. After filtration, cartridge filters were stored in an ice box to prevent degradation and later moved to the laboratory and kept at -20 °C in a refrigerator before extraction. DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen, Germany). Extracted DNA was stored at -20 °C and quantified using a Nanodrop 2000c (Thermo Fisher Scientific, Waltham, MA, USA) before sequencing.

To enhance the reliability of the environmental DNA survey findings, simultaneous observational surveys for fish and mammals were conducted. The traditional field survey, an indirect survey to identify mammalian habitats and traces of movement, was carried out by professionals in pairs on 21 July (summer) and 18 October (autumn) 2022. Summer and autumn were chosen as the survey periods to accurately ascertain the presence of otters in the target area, given that these seasons correspond to their low and high activity phases in South Korea, respectively (Kim 2018; Jo and Won 2019; Lee et al. 2020). Traces such as tunnels, footprints, dead bodies, food foraging, scat, ground nest and hair were investigated along the Saetgang. During the same period, fish were surveyed using a kick net, collecting fish for 40 minutes at each survey point and a casting net was employed 10 times. At the collection site, fish were identified to the species level, based on their morphological traits using Korean reference books (Kim 1997; Kim and Park 2002; Kim et al. 2005) and then released. This survey method is routinely utilised in the national natural environment survey in Korea (National Institute of Environmental Research 2016) In addition, a literature review preceded the fieldwork to determine the past presence of the Eurasian otter and other mammals in Yeouido between 2002 and 2019 (Gang et al. 2015; Seoul Metropolitan Government 2017; Yeongdeungpo Cultural Foundation 2020).

To determine vertebrate species and their taxon, DNA amplification was used together with three universal primers called MiFish (Miya et al. 2015), MiBird (Ushio et al. 2018) and MiMammal (Ushio et al. 2017). In order to create the Next Generation Sequencing (NGS) library, two phases of PCR were carried out. The first PCR was conducted with a total reaction volume of 12 μl, comprising 2 μl of DNA template, 6 μl of 2× KAPA HiFi ready mix (KAPA Biosystems, Inc., Wilmington, MA, USA), 0.36 μl of forward universal primer and 0.36 μl of reverse primer, with the remaining volume of 3.28 μl filled with distilled water. The three universal primers (MiFish, MiMammal, MiBird) were individually utilised in PCR reaction at a concentration of 10 µM. The primary PCR conditions were set to 35 cycles of 98 °C for 20 seconds of denaturation, 65 °C for 15 seconds of annealing, 72 °C for 15 seconds of extension and 72 °C for 5 minutes. PCR products were visualised on a 1% agarose gel to evaluate amplification success. Prior to the second PCR, the PCR product was purified using 20 μl of Ampure XP Beads. After purification, the concentration of the product was measured using the Infinite F200 PRO (Tecan) and normalised to 20 ng/μl. The reaction mixture for the second PCR was prepared with a total volume of 12 μl, consisting of 6 μl of 2× KAPA HiFi Ready Mix, 3 μl of distilled water, 1 μl of each unique dual index ID (P5 [Nextera, S5, 1 μM] and P7 [Nextera, N7XX, 1 μM], totalling 2 μl) and 1 μl of the normalised DNA template. The cycling conditions were identical to those of the first PCR, except for reducing the cycle number to 12. Following the second PCR, the concentration was measured again using the Infinite F200 PRO and adjusted to 100 ng/μl for sequencing. The Illumina libraries were sequenced using a MiSeq Reagent Kit v.3 (600-cycles) on an Illumina MiSeq 300PE platform (Illumina, San Diego, CA, USA). To address potential contamination, any amplification observed in negative controls may warrant exclusion of the detected organisms or the establishment of a threshold to filter out low-level noise (Forstchen 2020). Additionally, increased sequencing depth enhances the detection of rare species, with sequencing depths between 50,000 and 200,000 reads generally considered appropriate for metagenomic analyses (Bruce et al. 2021). Therefore, in this study, MiSeq sequencing results yielding fewer than 50,000 reads were considered negative, indicating a lack of amplification (Suppl. material 4).

The MiFish pipeline’s biogenetic information DB and the FASTQ file it analysed were compared (http://mitofish.aori.u-tokyo.ac.jp/mifish) to create a species list (Sato et al. 2018). For the bioinformatic processing of FASTQ files generated using MiBird and MiMammal, the PMiFish bioinformatics pipeline (PMiFish v. 2.4; the latest version is available from https://github.com/rogotoh/PMiFish.git; Miya et al. 2020) was utilised. Data preprocessing and analysis of MiSeq raw reads are performed using USEARCH v.10.0.240 (Edgar and Bateman 2010). The process begins with the assembly of quality-filtered forward (R1) and reverse (R2) reads. Subsequently, a length filter is applied to remove primer sequences and retain only those sequences that are not shorter than 140 bp. This is followed by quality filtering to further refine sequence integrity. After quality filtering, sequences undergo dereplication, where a depth filter is applied to ensure that only sequences with at least four replicates are retained. The next steps include denoising, to reduce error rates and taxonomic assignment of the cleaned and verified sequences. Results are rejected for sequences that are ambiguous or chimeric, with a BLASTn similarity of less than 97% to the clustered sequence.

Species list derived from metabarcoding results were refined in consideration of the study area (Gang et al. 2015; Seoul Metropolitan Government 2017; Yeongdeungpo Cultural Foundation 2020). The saltwater fish, alien species undocumented in South Korea and species extinct within native natural ecosystems were omitted from the detected species list (National Institute of Biological Resources 2021). Besides, considering the environment of the target site, some mammal detection results that do not display subspecies, such as wild boars (Sus scrofa) and wolves (Canis lupus), were recorded as domestic pigs (Sus scrofa domesticus) and dogs (Canis lupus familiaris), including subspecies. To ensure the accurate estimation of potential food species, we conducted a comparison between the species abundance, quantified by the number of individuals, obtained from traditional surveys and the species abundance measured in terms of the number of reads through eDNA metabarcoding. Before the analysis, the number of eDNA reads underwent a natural logarithmic transformation to standardise the data, consistent with the methodology employed in prior studies (Yates et al. 2019, 2021; Rourke et al. 2022).

A total of eight families and 13 species of mammals were recorded in the field survey and literature review (Table 1). From 2008 to 2019, there were no records of Eurasian otters in the study area. However, footprints of the Eurasian otter were observed twice during fieldwork for this study in October 2022. Given that the known home ranges of female and male otters are 7 km and 15 km, respectively (Néill et al. 2009), it is suggested that the otter can be detected across the study area, which spans 4.5 km. Additionally, mammals commonly found adjacent to urban areas, such as raccoon dogs, feral cats and rodents which were detected as the study sites, are within an urban area. Our observations and literature review did not reveal any presence of wild boars in the study area.

In our field study for fish, we identified a total of four families and 16 different fish species (Table 2). Fish abundance was notably higher at specific locations, particularly at the stream entrance (YS 1) and the trough (YS3), compared to other survey sites. There was a discrepancy in the number of species collected during April and October, but the overall species count remained consistent. Moreover, the autumn season exhibited a higher abundance of fish, with 381 individuals observed compared to 285 in the spring. Amongst the species observed, the number of Hemiculter leucisculus individuals increased from 72 in the spring to 148 in autumn.

The mammal list derived from metabarcoding results showed a total of six families and 10 species (Table 3). Negative samples, which were amplified to under 50,000 reads after MiSeq sequencing, indicated that there was no cross-contamination between samples. The eDNA survey revealed 6 of 13 species recorded in fieldworks and literature reviews, but there was no detection of Eurasian otters in both eDNA results (Fig. 2). Throughout surveys in April and October, Sus scrofa accounted for the highest percentage of reads (88.2%, 548,880 reads), followed by Bos taurus (6.7%, 41,745 reads), Ovis aries (2.9%, 18,078 reads) and Mus musculus (1.4%, 8,541 reads). It is estimated that genetic materials from domestic animals commonly consumed by people in the city were detected by the eDNA survey. In addition, the common urban species like the Muridae family were consistently identified by metabarcoding across all sampling sites and seasons.

Figure 2.

The comparison of identified mammal species between traditional survey and eDNA metabarcoding.

In contrast, the qPCR assay amplified Eurasian otters from the same samples analysed by metabarcoding. The LoD of the assay confirmed by the standard curve (Slope = -3.108, Y-intercept = 24.623, R² = 0.949, Eff% = 109.773), was determined to be 0.1 ng/µl DNA (Suppl. material 2). Positive samples were evaluated by the LoD calculated from the standard curve. Of the 10 PCR repetitions with 21 samples, five samples were positive for all 10 times, while nine samples were all negative. The trend line between the detection rate and Ct value showed was inversely proportional (Slope = -2.843 Y-intercept = 29.154, R² = 0.74) (Fig. 3), suggesting that the number of qPCR repetitions could be determined according to the concentration of samples and their detection rate. For instance, to achieve 100% of detection, the concentration of eDNA sample should exceed a cycle threshold of 26.3. With a high confidence level and considering the criteria, Eurasian otters were detected at three of seven sites (42.9%) in April (Fig. 4a). When thresholds were not considered, the number of detections increased to six of seven sites (85.7%). The repetition threshold with a high repetition number is more influential than the sample threshold in determining the number of detections. However, in October, despite differences in detection times amongst sample repetitions, eDNA detection confirmed Eurasian otters at all sampling sites (100.0%), regardless of the confidence level (Fig. 4b).

Figure 3.

The regression plot between the detection rate and a cycle of threshold (Ct).

Figure 4.

The number of detections according to criterion in April (a) and October (b). The reliability level was determined, based on the criterion that reliability increases when the threshold value is applied.

As Eurasian otters are known to prefer fish, making up over 70% of their diet (Suppl. material 5), only fish species identified in the eDNA metabarcoding results were considered potential food species in this study (Krawczyk et al. 2016; Narváez et al. 2020). A total of 3,898,508 and 1,089,451 reads, identified as fish species, were recorded in April and October, respectively. After excluding alien species not recorded in Korea and those unsuitable for freshwater, we retained 3,039,297 (Apr) and 748,001 (Oct) reads, constituting 78.0% and 68.7% of the raw data, respectively (Suppl. material 6). A total of 15 families and 52 species were detected. In both seasons, Cyprinidae accounted for the largest proportion of fish species at the family level (Apr, 54.57%; Oct, 43.58%), followed by Gobiidae (Apr, 16.90%; Oct, 22.88%) (Fig. 6). At the species level, Pseudorasbora parva was the dominant species in Saetgang, representing 5.68% and 6.35% of the natural log-transformed total number of reads (Suppl. material 7). Additionally, Tridentiger obscurus (5.16%), Acheilognathus macropterus (5.16%), Acheilognathus chankaensis (4.65%) and Cyprinus carpio (4.22%) were prominent in April. In October, A. chankaensis (5.05%) and T. obscurus (4.97%) were prevalent and Hemibarbus sp. (4.82%) and Zacco platypus (4.39%) also appeared. Similarly, in the conventional collection-based survey, Cyprinidae (Apr, 81.75%, Oct, 86.61%) and Gobiidae (Apr, 14.04%, Oct, 7.87%) were the dominant families (Table 2). At the species level, there was little difference in species with high relative abundance including H. leucisculus, but it displayed proportions similar to several other species, such as A. macropterus, C. carpio, P. parva and Z. platypus.

Figure 6.

The relative abundance (%) of detected fish species at the family level and species level in April and October a family level in April b family level in October c species level in April d species level in October.

Trace tracking survey and qPCR results confirmed the presence of Eurasian otters in Yeouido Ecological Park. Furthermore, the increased detection probability and the discovery of footprints in October suggest that Eurasian otters expand their home range during autumn. Eurasian otters are known to breed at times of the year when conditions are favourable, typically in early spring between February and March in Korea (Heggberget and Christensen 1994; National Institute of Biological Resources 2018). Since gestation lasts 60 to 70 days and weaning occurs at 3 months, it is estimated that the home range was reduced in April for nurturing purposes, coinciding with a decrease in detection probability (Han and Han 2022). Moreover, the concentration of DNA has been used in several studies to estimate biomass or population sizes (Takahara et al. 2012; Di Muri et al. 2020; Rourke et al. 2022). Therefore, the distribution patterns of Eurasian otters can be determined by the eDNA concentration further from the appearance location (Figs 5, 7). The eDNA survey conducted in October revealed that L. lutra was identified at all study sites. The eDNA concentration indicated that the density of L. lutra was higher at sites YS1 and YS2 compared to other sites, especially the lentic site YS3.

Figure 7.

Estimates of L. lutra density at the sampling site based on eDNA concentration in October.

As opportunistic predators, otters are likely to exploit locally abundant food sources, adapting their diet to seasonal variations and differences in habitat type (Ruiz-Olmo et al. 2001; Kloskowski et al. 2013). An analysis of prey preferences in Eurasian otters, based on spraint analysis from freshwater and estuarine environments in Korea, revealed that families Cyprinidae, Channidae and Centrarchidae exhibited the highest detection rates (Lee 2012; Kumari et al. 2019). In contrast, species from the orders Mugiliformes, Gadiformes and Pleuronectiformes were more prevalent in marine environments (Choi and Yoon 2012). Notably, the dominance of Cyprinidae, particularly P. parva, in April and October at this study area suggests a high availability of these species as a food source for Eurasian otters. Although environmental DNA surveys offer valuable insights into the habitat and ecological characteristics of species, it is essential to acknowledge and address potential biases that may influence these surveys.

In addition to implementing repeated site sampling and qPCR analysis in the lab to enhance the stability and reliability of PCR results, it is crucial to address challenges, such as the presence of inhibitors and primer efficiency in eDNA surveys (Macher et al. 2021). eDNA surveys enable highly sensitive detection of species that are low in abundance. However, the sensitivity of eDNA detection, especially with low sample concentrations, introduces uncertainties in presence-absence evaluations (Harper et al. 2018; Yu et al. 2022).

This study conducted three independent samplings at each site and ten independent qPCR analyses per sample. Findings showed that, in the 10 repetitions, 71.4% of the samples detected less than three times were considered positive, while those detected more than seven times accounted for the remaining 28.6%. This indicates the necessity of a minimum of three repeats to enhance detection probability. The decision-making process for establishing presence-absence criteria in qPCR assays can, therefore, be tailored, based on these results, with a high threshold for confident presence data or a lower threshold when determining broader habitat occupancy.

Although this study achieved high consistency in qPCR results without proceeding with inhibitor removal, eliminating or reducing inhibitors is vital for achieving high-quality detection results. Field methods, such as collecting small sampling volumes, using large pore-sized filters and increasing sampling intensity, help mitigate inhibitors (U.S. Fish and Wildlife Service 2022). However, these methods might reduce the concentration of target eDNA, necessitating careful application (Harper et al. 2019b). In the lab, commercial master mixes that include inhibitor removal columns and PCR enhancers provide additional tools to counteract inhibitors (Hunter et al. 2019; Sanches and Schreier 2020).

Additionally, understanding the characteristics of the study area, such as Yeouido Saetgang Ecological Park — an urban park vulnerable to genetic pollution from wastewater and artificial factors — is crucial before deciding on the analysis strategy (Lee et al. 2021). This study also found that otters were not detected and industrial and companion animals were frequently identified through eDNA metabarcoding. The study’s limitations due to cost and sample amount prevented repetition in metabarcoding, thus restricting comparisons between metabarcoding and qPCR detection rates. Moreover, the efficiency of the MiMammal universal primer was not evaluated against the Lutcyt species-specific primer for Eurasian otters. Although prior research has demonstrated that MiMammal primers can detect Eurasian otters (Ushio et al. 2017; Sales et al. 2020; Kim et al. 2021), further assessments focusing on comparative performance, environmental and sample type suitability and quantitative capabilities are required to improve result robustness (Bustin and Huggett 2017; Kumar et al. 2022).

Compared to other terrestrial mammals, semi-aquatic mammals spend most of their time in the aquatic environment engaging in activities, such as foraging and reproduction. This makes it easier to analyse environmental DNA (eDNA) through water samples. In other words, estimating the distribution and density of these species can be straightforward when obtaining eDNA samples with a certain concentration or higher. However, it is essential to consider the characteristics of the eDNA survey, including degradation, suspension and sedimentation, which can affect eDNA detection and concentration. eDNA particles share similar transport dynamics with fine particulate organic matter (Harrison et al. 2019). Lentic systems, such as lakes and ponds, are primarily influenced by gravitational settling and can disperse eDNA about 100 m (Nevers et al. 2020). Additionally, eDNA concentration in aquatic environments is commonly influenced or degraded by physiochemical factors, such as water turbidity, stratification caused by water temperature and benthic substrate (Deiner and Altermatt 2014; Shogren et al. 2017). Biological factors also play a crucial role. The production of eDNA varies not only with the biomass of the target species or community size, but also with species interactions like predation or competition (Stewart 2019). Therefore, the consideration of abiotic and biotic factors affecting eDNA detection should be reflected in the eDNA sampling and analysis strategy. Moreover, since the most reliable evaluation method involves assessing presence-absence through trace investigation or camera trapping and identifying the source of food through analysis of spraints and stomach contents, it is necessary to complement the eDNA investigation with these traditional methods.

The authors have declared that no competing interests exist.

No ethical statement was reported.

This work was supported by a grant from the Korea Environment Industry & Technology Institute (KEITI) through Creation Restoration and Management Technology of Carbon Accumulated Abandoned Paddy Wetland Project and funded by the Korea Ministry of Environment (MOE) (2022003630004).

Conceptualization: YK. Funding acquisition: YS. Investigation: YK. Supervision: YS. Writing - original draft: YK. Writing - review and editing: YS, SH.

Yujin Kang https://orcid.org/0000-0002-4904-8385

Seungwoo Han https://orcid.org/0000-0002-7148-4087

Youngkeun Song https://orcid.org/0000-0002-6996-3898

All the data supporting the findings of this study are available in the Suppl. material 6. Raw sequence data from the metabarcoding dataset are available in the Sequence Read Archive (SRA) under the accession number PRJNA1127811.