Diversity and site occupancy of fish communities
The identification of 31 taxa by eDNA metabarcoding represented more taxa than expected. Higher fish detectability by eDNA metabarcoding compared to conventional methods has been found in riverine as well as lacustrine environments (Hänfling et al., 2016; Li et al., 2019a; McDevitt et al., 2019; Olds et al., 2016; Pont et al., 2018). In each reservoir, some taxa were detected for the first time by eDNA metabarcoding, demonstrating the sensitivity of this method.
All conventional methods (i.e. boat electrofishing, fyke-nets, gillnetting and seine netting) are selective, thus there are several reasons for species underestimation with these tools (Kubečka et al., 2009). Small-bodied species, such as European bitterling (Rhodeus amarus ) in Klíčava and stone loach in Žlutice, were most probably overlooked in structured benthic habitats. Even species with medium body size, such as white bream in Klíčava and Žlutice, could be misclassified to the more common freshwater bream with similar appearance (especially juveniles) or disguised as hybrids (Kottelat & Freyhof, 2007). The other explanation is small population size in a water body and occurence in a specific habitat (Balasingham et al., 2017). Bullhead, Eurasian minnow (Phoxinus phoxinus ) and salmonids are restricted to relatively small, structured and shallow parts of tributaries, which are difficult habitats to sample by conventional methods. Typically netting is used, but net diameters, mesh sizes and used material determine the catch (Prchalová et al., 2009b; Říha et al., 2008). Large fish can escape from small nets or break through fragile nets, whereas small fish can escape through larger mesh sizes. Bleak, a common species in Římov, was probably recently introduced in Žlutice (maybe with piscivorous fish stocking), as the first record in gillnets was in 2019 (Blabolil,pers. obs. ).
eDNA metabarcoding is acknowledged as a highly sensitive method for detection of invasive and endangered species (Lawson Handley, 2015). Of the seven non-native species we detected, prussian carp and topmouth gudgeon have invasive potential in the Czech Republic (Pergl et al., 2016). Spread of these species can be facilitated by outflowing water from ponds and therefore early and reliable detection is imperative to help protect native fishes. Pumpkinseed was detected only in main part of Římov and will require immediate action to prevent further spread as it is naturally reproducing in the reservoir. Grass carp (Ctenopharyngodon idella ), Hypophthalmichthys spp., rainbow trout (Oncorhynchus mykiss ) and Salvelinus spp. are dependent on artificial stocking, but grass carp andHypophthalmichthys spp. are undesirable in reservoirs. Grass carp reduce submerged vegetation and Hypophthalmichthys spp. filter zooplancton in the water column, supporting phytoplankton. Therefore, early eDNA detection of undesirable species should trigger their elimination, similar to eradication management of common carp (Cyprinus carpio ) in Tasmania (Furlan et al., 2019) and topmouth gudgeon in the UK (Robinson et al., 2019).
Three taxa were classified as regionally Extinct as current populations are dependent on artificial stocking. Common carp is the most popular game fish in the Czech Republic and stocked in ponds across the country (Boukal et al., 2012). Ponds in catchments are the most probable source of individuals found in the studied reservoirs. Coregonus spp. and Acipenser spp. were detected only in Římov. Ponds occur in this catchment as well as fish hatcheries breeding this species (Pstruhařství ČRS Kaplice, Ltd.) that are directly connected to the inflowing river. Potential escapees from pools is therefore plausible (Naylor et al., 2001). The European eel is categorised as Extinct in the Wild because migration of juveniles upstream is impossible and high mortality of adults during downstream spawning migration through dams with hydro turbines (Trancart et al., 2019). European eel eDNA detection in Klíčava and Římov could help initiate construction of suitable dam bypasses and/or capture of spawning stock and transport downstream past the dam.
The crucian carp (Carassius carassius ) was the only Critically Endangered species we detected and typically occurs in small ponds (Harper et al., 2019a), but here was exclusively detected from the side inflow to Římov. In this particular case, it is very probable that eDNA originated upstream in a pool, the closest one being around 670 m from the sampling locality. This suggests that populations of endangered species could be identified through eDNA transport. Vulnerable taxa are rare in the reservoirs associated with tributaries (Eurasian minnow, grayling and Lampetra spp.) or submerged macrophytes (tench [Tinca tinca ]), but these species can be considered as indicators of ecological quality and diverse habitats (Blabolil et al., 2017b, Šmejkal et al., 2014). The category of Near Threatened species is important as these species could be at high risk of extinction in the near future and therefore population assessment should follow any eDNA detections. Finally, most of the species we detected were classified as Least Concern, suggesting that conditions were suitable for common species.
The cumulative number of detected taxa was highest (29) in Římov (largest reservoir) and lowest (16) in Klíčava (smallest reservoir). Our observation confirms a general increase of fish diversity with the area, volume and catchment size of a water body (Mehner et al., 2005). However, the number of taxa in individual samples was much lower, typically in single units. This confirms the observation from previous studies that eDNA is not homogeneously distributed in lentic habitats and reflects local species presence (Harper et al., 2019a; Lawson Handley et al., 2019; Li et al., 2019b). We observed greater taxon detection in autumn compared to summer, which may be due to prolonged eDNA persistence at colder temperatures (Collins et al., 2018; Lawson Handley et al., 2019). Summer sampling is more likely to reflect contemporary species distribution and activity compared to colder conditions that preserve eDNA and facilitate transport, and could provide more accurate information on changes in species activity (e.g. downstream migration for wintering, Jonsson & Gravem, 1985) as well as eDNA flushed from upstream. These results provide further evidence that spatial and temporal replication in eDNA surveys is required for reliable estimates of species richness (de Souza et al., 2016; Djurhuus et al., 2020; Hänfling et al., 2016; Lawson-Handley et al., 2019; Li et al., 2019b).
Despite this spatiotemporal heterogeneity, we found semi-quantitative abundance estimates from eDNA metabarcoding in reservoirs (i.e. site occupancy) were highly correlated with abundance categories based on long-term fish surveys, which corroborates previous studies in UK lakes (Hänfling et al., 2016; Lawson Handley et al., 2019). The majority of fish in our studied reservoirs were of riverine origin, but the most plastic species, such as freshwater bream and roach, prevailed and occupied new niches of more or less standing water (Blabolil et al., 2016). Although the pelagic zone was more homogeneous compared to the structured littoral zone with different bottom substrate, vegetation, slope, Bray-Curtis dissimilarity values were lower (indicating higher similarity) for littoral samples compared to pelagic samples. This may be due to the lower number of taxa per sample in pelagic habitats resulting in more dissimilar samples. Studies in other geographic regions also found more species in the littoral compared to the pelagic zone (Lawson Handley et al., 2019; Zhang et al., 2019). Overall, more taxa per sample resulted in higher similarity as shown in tributary samples.