A comparison of European eel Anguilla anguilla eDNA concentrations to fyke net catches in five Irish lakes

The critically endangered European eel Anguilla anguilla (Jacoby & Gollock, 2014) inhabits inland and coastal waters from northern Norway throughout the Mediterranean as far south as the north-western coast of Africa (Dekker, 2000, 2003). Reports of drastic population decline across the whole region are well documented, with abundance currently estimated at 5% of that observed in the 1980s (Dekker, 2004, 2016; Henderson, Plenty, Newton, & Bird, 2012; MacNamara, McCarthy, & Barry, 2016; Stone, 2003). In parts of Europe and East Asia, eels are still considered to be a rare and luxury food item. European and other eel species cannot yet be captive bred to maturity (Butts, Sørensen, Politis, & Tomkiewicz, 2016), so consequently fishing that removes Received: 20 September 2019 | Revised: 28 April 2020 | Accepted: 30 April 2020 DOI: 10.1002/edn3.91

individuals directly depletes the remaining breeding populations.
Wild-caught glass eels (juvenile) and elvers (pigmented juvenile) are in demand to stock commercial eel farms that "grow on" eels to supply the market (Sustainable Eel Group, 2018). This, coupled with their scarcity, has ensured that they maintain a high monetary value and these fish are among the most trafficked critically endangered fish species in Europe. The UK based Sustainable Eel Group estimates that 100 tonnes of glass eels are exported from Europe to Asia annually; the final eel fillets produced from these glass eels will have a value of €2.27 billion at the point of consumption (Stein, 2018).
Legislation exists to enhance the recovery of A. anguilla populations (European Union Eel Regulation, 2007) which is interpreted and implemented in each EU country through Eel Management Plans. It is recommended that, in order to conserve A. anguilla numbers, at least 40% of mature eels should leave European waters to migrate to spawning sites. To meet the goals of the Eel Management Plans and ensure these levels are reached, all life stages of A. anguilla residing in European waters, need to be accurately monitored.
Nearly, all watercourses in Ireland have a population of A. anguilla with the large lake-wetted areas that are available being the preferred habitat for eel growth (Moriarty, 2003). Once a thriving eel fishery, fishing for A. anguilla was suspended in Ireland in 2009 (Conservation of eel fishing, bye-law number C.S. 303) through its Eel Management Plan (Inland Fisheries Division, 2008).
Consequently, the only fishing for A. anguilla is now exclusively for monitoring purposes. The Irish agency, Inland Fisheries Ireland (IFI), has a dedicated eel monitoring team and during the summer months (June-September) conducts fyke-net surveys in freshwater lakes to measure eel abundance and health.
Fyke nets are a standard research method for capturing benthic fish species such as eels and have been used in Ireland to study eel populations since 1965 (Moriarty, 1975). They have a large capture capacity and reduced mortality rate compared with other netting techniques (Krueger, Hubert, & Price, 1998). However, despite their widespread use they are difficult to set correctly, require considerable handling of the eels, and in low eel populations may underestimate the total number of eels (Jellyman & Graynoth, 2005). A more sensitive noninvasive detection method is required, particularly for declining eel populations when eels are scarce.
Fish and other organisms leave behind DNA (through skin cells, scales, feces, and mucus) as they move through their environment, which if sampled before it degrades or settles into the sediments can be a useful indication of their recent presence. This shed DNA is more commonly referred to as environmental DNA (eDNA). Methods that detect eDNA from aquatic environments are frequently found to be more sensitive than traditional presence/absence methods of detection (Sengupta et al., 2019;Smart, Tingley, Weeks, Van Rooyen, & Mccarthy, 2015) and can be powerful tools for the ecological monitoring of aquatic organisms. eDNA has already been used to detect single species and methods have been described for many aquatic organisms (Agersnap et al., 2017;Harper et al., 2019;Rees, Bishop, et al., 2014;Sigsgaard, Carl, Møller, & Thomsen, 2015). To advance eDNA for routine and reliable monitoring, methods need to undergo robust validation and demonstrate their ease of use. They should also reduce labor requirements and increase sensitivity over traditional survey methods (Goldberg et al., 2016).
However, to monitor A. anguilla in freshwater lakes this study designed and validated a qPCR method in order to directly compare eDNA monitoring to the results of fyke net fishing for A. anguilla.
Other comparative eDNA studies have been found to be more sensitive than fyke net fishing when detecting the presence or absence of common carp (Cyprinus carpio), red fin perch (Perca fluviatilis) and Oriental weatherloach (Misgurnus anguillicaudatus; Hinlo, Furlan, Suitor, & Gleeson, 2017) and crucian carp (Carassius carassius; Harper et al., 2019). A recent study that compared eDNA monitoring with electrofishing for Japanese eels Anguilla japonica in rivers, also found eDNA methods to be more sensitive at detecting riverine distribution of eels (Itakura et al., 2019).
This study focussed on A. anguilla in Irish lakes. The study compared a single species eDNA method for the detection of A. anguilla with paired fyke net surveys conducted by the IFI eel monitoring team during summer 2018. The method was developed with compliance to the minimum information for publication of quantitative real-time PCR experiments, MIQE guidelines (Bustin et al., 2009).
The field validation of this eDNA method is described using filtered surface water samples collected from five freshwater lakes across Ireland. These lakes are classed as having high to low eel populations based on historic fyke net data. The potential application of monitoring eel populations in lakes using eDNA is discussed.

| Study species
The European eel, A. anguilla, is a migratory fish that spawns in the Sargasso sea (van Ginneken & Maes, 2005). The larval eels (leptocephali) are thought to drift on ocean currents toward European waters, a journey that can take between a year and maybe more than eighteen months (Bonhommeau, Castonguay, Rivot, Sabatié, & Le Pape, 2010;Friedland, Miller, & Knights, 2007;Lecomte-Finiger, 1994;Zenimoto, Sasai, Sasaki, & Kimura, 2011). Here, environmental triggers initiate the transformation of the larvae into glass eels (up to 5 cm long and transparent), which move into coastal estuaries and migrate upstream into freshwater becoming pigmented juveniles (elvers) and developing into the yellow eel (growth stage; van Ginneken & Maes, 2005).
The pigmented yellow eels (growth stage) may spend 6-22 years here in a benthic sedentary stage, feeding, and growing until they start to reach sexual maturity (Arai, Kotake, & McCarthy, 2006). Once mature, the adult silver eels (reproductive stage) will move back downstream into coastal waters and commence the lengthy migration back to the North Atlantic spawning sites. These migratory events are seasonal, with glass eels arriving in Ireland in significant numbers between January and May and the mature silver eels leaving between August and December (Russell Poole & Reynolds, 1998).

| Location of study
A number of lakes were scheduled for monitoring in 2018 as part of the larger IFI eel monitoring program and out of these lakes, six were chosen for this eDNA analysis. The lakes sampled for the study were categorized as high, medium, and low/absent eel populations.
Unfortunately, Lower Corrib, which would have paired with White Lough (a medium population lake), was canceled due to a bereavement in the local fishing crew leaving just five lakes for the analysis (Table 1). The eel population designation is simple due to the variation in eel catches over time and the size of lake. Simple categories have been used in other eDNA detection studies (Ficetola, Miaud, Pompanon, & Tab erlet, 2008). Historically, Lough Derg (Shannon catchment) and Lough Corrib (Corrib catchment) supported a large commercial eel fishery and together with survey data indicated a large eel population. Lough Owel (Shannon catchment) was also the site of a commercial fishery but small in scale, and recent fyke surveys already indicated a declining population. Lough Muckanagh Lough are in the upper reaches. The location of the lake within a catchment does impact on the distribution of eels (Degerman, Tamario, Watz, Nilsson, & Calles, 2019) but the focus of this study was designed to test the eDNA method with specific levels of eel abundance (high, medium, and low). Within all lakes, the net depths were set <20 m, with the majority set <5 m ( Table 1). Analysis of eDNA was undertaken in the lowest population lakes (Lough Owel and Lough Muckanagh) to determine whether eDNA sampling for A. anguilla is more sensitive than fyke netting.

| Field sampling
Sampling and filtration took place on fifteen days between June and September during 2018. Water samples were collected using a TA B L E 1 Site descriptions and water samples collected for eDNA analysis and results boat on twelve days, at fyke-net locations before and after fishing

Lake
and at independent open water lake locations not associated with fishing sites. An additional three days of water sampling was conducted from lake shores. In total, including the filtration controls, 110 water samples were filtered for eDNA analysis. Across all five lakes, 1 L samples were collected from 21 fyke net locations. Water samples were collected before nets were deployed (referred to as presamples) and again above the fyke nets 18-24 hr later (n = 42) before the nets were lifted (referred to as postsamples). At each lake, water samples were also taken at shore Poor weather precluded boat-based sampling on days 3 and 4, so 2 L water samples were collected on foot from accessible points along the shore-side at the same locations on both days. Sample locations can be found in Figure 1. Temporal autocorrelation is highly unlikely to be a significant factor in the data as eDNA concentrations are not expected to change significantly within the water column over a 24 hr time horizon.
Sample collection was standardized for both one-and two-liter water samples. Wearing clean disposable nitrile gloves, the sample bottle was briefly rinsed using water from the lake. The water bottle was immersed at the water surface until full. The bottle was labeled with the time of collection and the sample location recorded using GPS. Where water samples were collected from the shore-side specific care was taken to avoid disturbing lake sediments.
Filtering of water samples was conducted at temporary field laboratories within eight hours of collection. A Masterflex tm peristaltic pump head (Cole-Parmer) was operated using a generic cordless drill. This drew a single water sample via silicone tubing through a closed in-line filter holder (Sarstedt) containing a 49 mm cellulose nitrate 3 µm pore filter (Whatman ® ). Wearing clean gloves and using bleach-treated forceps, each filter was transferred into a 2 ml screw F I G U R E 1 Map of Ireland with the locations of the five freshwater lakes sampled. Within each lake, the pie charts indicate the location of 1 L water sample locations. Positive Anguilla anguilla eDNA PCR replicates (n = 6) are in blue and negative are white for each water sample. The labels, shore, open, pre and post describe the type of sample collection, shore, and open lake are non-net associated, pre and post are collected at the fyke net locations before the nets are set and the following day before the nets are lifted. Numbers on the postnet samples indicate the number of eels caught cap sample vial that contained 800 µl of Longmire's storage buffer (Williams, Huyvaert, & Piaggio, 2016). Occasionally suspended solids blocked the filter membrane, when this occurred it was replaced with a second filter and the remainder of the water sample filtered.
The two filter papers were rolled together and placed into a single vial thus all filtered samples were stored in the same standard volume of buffer. On no occasion were more than two filters required.
After each filtration procedure, the filter housing and tubing were completely submerged in a 10% bleach (sodium hypochlorite) solution for 20 min, then transferred and completely submerged in cold tap water to remove the bleach. Each filtering batch included a filtration control of 1 L tap water (filtration negative control). The labeled vials with the filter paper and Longmire's buffer were stored at ambient temperature and transferred to the laboratories at the University of the West of England, Bristol (UWE) for analysis.

| eDNA extraction
Total DNA was recovered from the filter and buffer using a phenol

| Design of species-specific qPCR primers and probes
Anguilla anguilla mitochondrial cytochrome b sequences were obtained from NCBI Genbank and aligned using MEGA version 10.0.4 (Kumar, Stecher, Li, Knyaz, & Tamura, 2018). Species-specific primers and probe were identified using the NCBI primer-BLAST feature (http://www.ncbi.nlm.nih.gov/tools /prime r-blast /) and optimized visually for qPCR. Specificity of the primers was determined in silico using the primer-BLAST tool. This showed no risk of cross-amplification with nontarget species and closely related Anguillid species.
The primers' specificity was next challenged in vitro by amplification with genomic DNA extracted from UK and Irish fish species likely to be encountered during sampling and DNA from the closely related freshwater eel species Anguilla rostrata (Table S1). The third specificity check was a confirmation sequencing step of environmental-derived positive PCR products to establish that the correct A. anguilla cytochrome b fragment was being amplified.

| Quantitative real-time PCR
All of the DNA extractions were assayed for A. anguilla eDNA in a qPCR, using specific primers and a hydrolysis probe on a StepOne Plus TM Real-Time PCR System (Thermo Fisher Scientific). Plates were prepared in a clean laboratory using a dedicated PCR workstation previously decontaminated using UV light for 30 min. The limit of detection (LOD) for the qPCR primer-probe set was established using a standard method previously described by Broeders et al. (2014).
Briefly, a serial dilution series was prepared of a 998 base pair synthetic gene copy of the A. anguilla cytochrome b gene (Integrated DNA Technologies) from which the exact copy number of qPCR amplicons was determined. Using this, two forms of LOD were calculated. The LOD 6 , which is determined by identifying the last dilution in a series of sixfold dilutions where all six PCR replicates give rise to specific and positive amplification. The second the LOD 95% determines the specific sensitivity of the primer set with a 95% confidence level. This was obtained by amplification of 60 replicates of the lowest serial dilutions around the LOD 6 . For this primer-probe set, both LOD 6 and LOD 95% was twelve haploid genome equivalents per reaction. During this study, the total catch from each net was returned to shore where the fish were weighed and counted before being returned to the lake ( Table 2). All eels caught were yellow eels (growth stage) specimens. A proportion of the catch were randomly selected by IFI to collect morphometric data. This included total length (to nearest mm), weight (to nearest g), horizontal eye diameter (mm), vertical eye diameter (mm), pectoral fin length (mm), diameter of head and pigmentation (data not presented). These eels were then released back to the lake.

| Data analysis
Statistical relationships were analyzed using Gaussian generalized linear mixed models with identity links. During analysis, residuals from each model were visually inspected to ensure normality, homogeneity of variance and linearity. The relationship between eel biomass in individual fyke nets and mean eDNA concentration in the paired postnet samples was modeled using biomass as a covariate and lake as a random factor. To determine whether eDNA concentrations predict overall biomass within individual lakes, a critical question for fisheries management, the mean eel biomass was calculated from all fyke nets for each lake. This was included as a covariate nested within lake, with lake being included as a random factor. Variation between the con-  (1 vs 2 L samples) on concentrations of DNA was modeled separately using volume filtered as a fixed factor and sample as a random factor.
All analysis was carried out using the lme4 package (

| eDNA sampling in freshwater lakes identifies the presence of European eels
The data demonstrate that A. anguilla can be detected in Irish lakes using this single species eDNA method. A. anguilla eDNA was de- In the laboratory analysis, all extraction and PCR negative controls tested negative with no evidence of contamination. The PCR efficiency ranged from 94.04% to 99.67%, the slope ranged from −3.33 to −3.53 and the y-intercept from 22.36 to 27.36. The R 2 values ranged from 0.98 to 0.99. Samples that gave rise to a negative result were tested by re-amplification with 1 µl 0.03 ng/ml A. anguilla genomic DNA. This indicated that failure to amplify was not due to environmental inhibition, and that these samples were likely to be either negative or below the detection limits (LOD 6 and LOD 95% ) of the qPCR for field samples of A. anguilla eDNA.

| The eDNA concentrations at net locations reflected the eel biomass recovered in fyke nets
The abundance of A. anguilla was established by eel catches in the fyke nets deployed, 67% (14/ 21) of nets deployed caught eels, all growth stage specimens. All eel catches were from lakes histori-

| Lake eDNA concentrations predict average biomass
The influence of the A. anguilla biomass and sample location (shoreside vs open water) were analyzed. The eDNA concentrations predicted the average biomass retrieved from each lake (F 1,3 = 20.60, p = .020, r 2 = .425; Figure 3). Furthermore, the eDNA did not differ significantly between samples collected shore-side and from open water (F 1,3 = 1.60, p = .295; Figure 4).

F I G U R E 2
Anguilla anguilla eDNA concentration in 1 L postnet samples and the biomass (kg) of eels subsequently recovered by fyke net at those net locations (n = 18). White Lough postnet samples and catch data were removed from the dataset due to detection of eDNA in the filtration control. Note the log scale on the y-axis and two data points of low eel catches with no eDNA detected. IFI categorizations of eel population size in each lake are given in brackets At two lakes, Lough Owel and Lough Muckanagh, no eels were caught using fyke nets during this monitoring period (Table 2).
However, A. anguilla eDNA was detected in 91% (n = 23) of all the 1 L water samples (including shore-side and open water, preand postnetting) collected from these lakes (Figures 1 and 3). The eDNA quantities recovered in these 1 L water samples from low eel F I G U R E 3 Quantity of eDNA recovered by location for each lake. Each data point represents a water sample from which Anguilla anguilla eDNA was recovered. IFI categorizations of eel population size in each lake are given in brackets F I G U R E 4 Comparison of eDNA quantities (log-transformed) in 1 L water samples (n = 40) collected from shoreside (n = 11) and open water (non-net associated and prenet, n = 29) from all five lakes. The eDNA quantities are logtransformed for visual clarity, the box plot represents quartile range of the data F I G U R E 5 Quantity of eDNA in 1 L samples from shore-side and open water samples shown at all five lakes. The box plots describe log-transformed data represented as quartile ranges. Here, * is more than three times the interquartile range (extreme outlier), and o is more than 1.5 times the interquartile range (outlier). IFI categorizations of eel population size in each lake are given in brackets populations, was much lower than in the other lakes ( Figure 5). Fyke netting at Lough Muckanagh subsequent to this study confirmed eel presence within the lake.

| Sample volume
We tested whether the volume of water filtered influenced the concentrations of DNA using 1 and 2 L samples taken at Upper Corrib.
While there was more eDNA in the 2 L samples than the 1 L samples where eDNA detection occurred there was no significant difference in the quantity of eDNA recovered per L between 1 and 2 L samples (F 1,25 = 0.65, p = .428, r 2 = 0.013; Figure S1).

| Contamination control
A potential challenge with eDNA sampling is cross-contamination of the samples. Extensive contamination controls were employed at each step of the eDNA analysis which were negative for all laboratory steps. All of the sample controls (filtration; DNA extraction; PCR amplification) from the low population lakes were negative. Field contamination occurred on one occasion (White Lough, discussed further below). All samples associated with this filtration batch were removed from the dataset.

| The detection of Anguilla anguilla eDNA in freshwater lakes
The primers and probe designed for this study were able to detect eDNA of A. anguilla origin in all five freshwater Irish lakes. Although the detection of A. anguilla eDNA has been previously reported in single species and metabarcoding detection studies (Seymour et al., 2018;Takeuchi et al., 2019;Thomsen, Kielgast, Iversen, Møller, et al., 2012), it was not applied for field monitoring. There is signifi-  (Table   S1). However, two additional samples from eels identified as A. rostrata caught in a European lake (most likely derived from a restocking event) did weakly amplify using these primers and probe.
For this study, we relied in part, upon the natural geographical distribution of these two species, with the chance of encountering naturally occurring A. rostrata in the Irish lakes sampled being highly.
Globally, the movement of anguillid species is a concern and there are reports of A. rostrata in Europe (Frankowski et al., 2009). It is suspected that A. rostrata located in Europe are introductions and a result of imported eels for stocking and aquaculture. Ireland has never imported glass eels as part of a stocking program. Large A. anguilla stocking programs have only been carried out in two catchments in Ireland, the Shannon, and the Erne system. All stocking that was carried out in these rivers was the movement of locally sourced glass eels from below the hydrostations, supplemented with catches from neighboring catchments but only in the River Shannon. Therefore, the risk of A. rostrata being present in Ireland is very low.

| Using eDNA to monitor Anguilla anguilla in freshwater lakes
In this study, five freshwater lakes in Ireland were simultaneously monitored for European eels using fyke nets and a single species eDNA technique. The sampling took place between June and September 2018, in freshwater lakes categorized by IFI into high, medium, and low eel populations based on historical catch data. This study shows that A. anguilla eDNA can be successfully recovered and amplified from freshwater lakes (Figure 1). There was a clear relationship between the overall quantity of eDNA recovered and eel biomass in the lakes. The eDNA concentrations recovered from the five lakes compared well to their high and low eel population descriptions.
Simple sampling from surface waters is another positive aspect for applying eDNA methods for monitoring. Taking samples at the surface are logistically an easier method than taking samples at depth. Collecting samples at depth requires specialist equipment and disturbing the lake sediments must be avoided which can be difficult to control for from a small boat.  looked at the DNA copy number of benthic weatherloach and found no effects sampling water at the surface or subsurface. Furthermore, other studies of benthic species have collected samples taken at the water surface (Forsström & Vasemägi, 2016;Wilson et al., 2014) and successfully detected the target species. Notably, in order to standardize eDNA sampling during this study water samples were always Windemere during the summer months than in Winter. In addition, yellow eels (growth-phase) are demersal fish, and generally found in the lowest benthic region of rivers or lakes (Capoccioni, Lin, Iizuka, Tzeng, & Ciccotti, 2014). Taken together, it would be reasonable to suggest that eDNA sampling for A. anguilla should be from lower lake depths. Unless that is, the lakes are shallow and mixing of the water column is occurring. In this study, the majority of nets were set in shallow water (<5 m) in Lough Derg, Lough Muckanagh, Lough Owel, and White Lough. The eDNA comparisons with the nets set in these shallow lakes are good, it may be because there are unlikely to be strong lake thermoclines at these depths. The nets in Lough Corrib were a mix of depths but all less than 20 m. By sampling DNA at the surface in the summer, our results should not necessarily link to the net depths at 20 m (especially as it was in the summer). The fact they do potentially demonstrates that a seasonal thermocline was not limiting for this surface eDNA surveys for demersal species, including eels. With the nets set at different depths (by the nature of the lake floor) both across a lake and between lakes, standardizing the eDNA sampling at the water surface also reduced the variation between samples.
In the first week of sampling, both 1 and 2 L water samples were collected for method validation. In the two low eel population lakes, 91% of the water samples collected were positive for A. anguilla eDNA. These lakes were included in this study specifically to determine if eDNA sampling may be more sensitive than fyke netting especially if catches are low and infrequent. Traditionally, both lakes had been commercially fished for A. anguilla but long-term monitoring by IFI over the past 10 years has indicated that eel populations are declining in these small upstream lakes, with very low numbers of eels recorded in recent surveys (Table 1). In 2002, Ibbotson, Smith, Scarlett, and Aprhamian (2002) predicted that the decline in recruitment of A. anguilla would lead to a decrease in the proportion of eels moving into these freshwater habitats. Although fyke net sampling failed to catch any eels in these lakes (during the eDNA survey), both the pre-and post-eDNA samples indicated the recent presence of eels at multiple locations around the lakes. Furthermore, the positive A. anguilla eDNA results were confirmed at Lough Muckanagh in a repeat fyke net survey.
We recovered eDNA from these lakes with a high degree of repeatability suggesting that eDNA can be more sensitive for assessing species presence than fyke net sampling when populations are low.
It was a difficult undertaking to conduct the eDNA sampling alongside fyke net fishing and a testament to the careful adherence to collection methods that cross contamination occurred only once during the study. The eDNA methods are highly sensitive, in the laboratory under optimal LOD conditions the primer and probe set are able to amplify as few as twelve haploid genome equivalent copies of target mitochondrial DNA.

| Variation in eDNA concentrations between lakes
Although the mean eDNA quantities differed between the lakes, within each lake the eDNA quantities did not significantly differ between samples collected either at the shore-side or from the open water locations ( Figure 5). As well, the collective relative eDNA concentrations for each lake compared well to the fyke net categories (Figures 3 and 5). This raises two points concerning the distribution of eels within each lake and category (high, medium, and low). If the eel density across the lakes is evenly distributed, then simple shore side sampling is sufficient to inform the lakes general population, potentially a key finding. However, if eel density is not evenly distributed then this survey strategy was not reflective of local differences in the numbers of eels.
Exactly, how eels behave in lakes is still poorly understood, different sampling methods conclude differing distributions.
Using a baited longline, Yokouchi, Aoyama, Miller, McCarthy, & Tsukamoto, 2009, found the eel density (CPUE) in a lake in Ireland was highest in the deepest water (open water) than shallow water (shore-side). Conversely, at Upper Corrib during this (unbaited) fyke net survey the highest number and biomass of eels were caught in the shallowest nets set at 0.5 m, and no eels were caught in the deepest nets set at 16 m ( Figure S2). To understand the distribution of eels in lakes through an eDNA survey, it is likely that a completely different sampling strategy is required. The data hint that it may be possible to reflect local eel abundance within each lake. Figure 2 shows the eDNA concentrations from samples collected above nets with eels and suggests eDNA concentrations do reflect the number of eels contained below. However, samples collected at net locations before the nets were set did not always predict subsequent eel catch (Figure 1). Most likely other factors are having an effect that were not measured during this study and an intensive standardized sampling effort would make a useful future study-again paired with comprehensive capture methods. Determining a link between eDNA concentrations and eel abundance in the field will ultimately allow a more accurate assessment of eel presence and abundance.
The aim for this study was to confirm the presence or absence of eels in freshwater lakes. In addition, the data confirmed the comparative population status (determined through fyke net surveys) of the lakes through relative eDNA concentrations in the water samples.
Both survey methods (fyke nets and eDNA) provide valuable and complementary information and together can provide a robust sampling regime. This is particularly important as the distribution and recruitment of A. anguilla is so low (Dekker, 2004;Henderson et al., 2012).

| Method validation to inform management practise
For eel management programs, using fyke net catch data alone can be problematic. The fyke net is a biased sampling method, the size of the eels caught and retained is determined by the mesh size of the cod end (usually eels larger than 25 cm). It is only the eels caught that are over 30 cm where 100% of the catch is retained (Bevacqua, De Leo, Gatto, & Melià, 2009). In this study, the fyke cod-end mesh size was 12 µm and the eels captured ranged from 26.5 to 86.9 cm (Table 2). If eel numbers are low and netting is infrequent the survey may fail to catch any eels, leading to a false zero (Martin et al., 2005) or when the distribution of eels is patchy within a lake this can result in highly variable catches (Rose & Kulka, 1999).
In low eel population lakes, where fyke net surveys no longer catch eels, switching to eDNA methods has the potential to confirm local extinction or to detect an upturn and re-establishment of a population. This will be particularly useful for habitat remediation works and barrier mitigation where assessment of recruitment of this critically endangered species without using invasive capture measures will be valuable.
In conclusion, this study compared single species A. anguilla eDNA survey method to a fyke net survey. The eDNA method is quick and easy to perform in the field and samples can be collected from the shore as well as from a boat by staff with minimal training.
Sampling could be conducted shore-side and was more sensitive than fyke nets in detecting low eel populations. The increased sensitivity of eDNA methods will be a useful additional tool to guide survey design and effort. Furthermore, repeated use of eDNA surveys over seasons would build a clearer picture of changing eel abundance to compliment fyke net catch data. While the eDNA method cannot replace the fyke net surveys, this method is a more sensitive indicator of eel presence and particularly useful for monitoring population advances or retreats within freshwater systems and for enabling rapid assessments of populations across more locations than is possible using live-capture techniques.

ACK N OWLED G M ENTS
This study was part of the PhD project of LW, funded by the like to thank the two anonymous referees for their valuable comments on a draft of this manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
LW, CO, MS, HM, LN, and SS conceived the study and designed the methodology; LW and CO collected the data; LW, MS, and LN analyzed the data; LW led the writing of the article. All authors contributed critically to the drafts and gave final approval for publication.

DATA AVA I L A B I L I T Y S TAT E M E N T
The sequence data generated for this study are available in the supplementary data. The eDNA dataset analyzed during the current study is available from the University of the West of England data repository http://resea rchda ta.uwe.ac.uk/571.