Environmental correlates of fish distribution
Interpretation of eDNA-based results is not straightforward as eDNA dynamics, including deposition, accumulation, transport, and fate, and fish distribution are influenced by biological and environmental variables (Barnes et al., 2014; de Souza et al., 2016; Mehner et al., 2005). In heterogeneous environments, hydrodynamics affecting water movements should be taken into account. Flow in lotic environments, such as rivers or brooks, combines eDNA of local residents with eDNA being transported from upstream (Jane et al., 2015; Pont et al., 2018). Flow is also present in reservoirs and most likely explains the combined detection of taxa associated with running water and upstream standing waters as well as the high diversity found in the main inflows of the study reservoirs. For example, detection of common carp and topmouth gudgeon in running water may have been due to eDNA transport from commercial ponds in the catchment, although escaped fish from these locations could also explain the eDNA signal in rivers. Downstream eDNA transport of eDNA was further supported by the negative correlation with normalised water age for some taxa. In contrast, a positive correlation with normalised water age reflects species’ preference for a particular habitat and lentic conditions, e.g. prussian carp and European bitterling (Kottelat & Freyhof, 2007).
Wind and wave action can also create currents in lentic environments that transport eDNA (Moyer et al., 2014). This effect is more pronounced in summer, with strong thermal stratification separating layers of different temperatures and densities. In this study, species associated with reservoir bottoms, such as European eel, stone loach and barbel (Barbus barbus ), were detected only in littoral habitats. One exception is ruffe, the most common species associated with the benthos in the studied reservoirs that has a pelagic juvenile phase before settlement on the bottom (Jůza et al., 2015). Ruffe detection in both habitats can be explained by potential horizontal currents as well as eDNA shed by juveniles. Nonetheless, horizontal eDNA transport could prove useful to detect different ontogenetic phases of a species with eDNA metabarcoding if they utilise distinct habitats (Zhang et al., 2019).
Despite eDNA transport facilitated by flow, wind, and wave action in reservoirs, eDNA signals for different taxa were spatially structured and reflected species’ ecological niches. Spatial distribution of fishes in heterogeneous freshwater bodies can result from various environmental factors individually or combined, but the most important factor is temperature. Higher temperature stimulates metabolic and behavioural activity of most poikilotherm fish (Stoner, 2004), which can also increase eDNA release and deposition by fish into environment as they exhibit greater rates of movement, feeding, digestion, and excretion (Jo et al., 2019; Robson et al., 2016). More active fish may have higher eDNA detection probabilities; however, warmer water temperature is also associated with greater UV irradiation, greater microbial activity (including feeding on eDNA as a dissolved organic phosphorus substrate), and reduced oxygen solubility (Barnes et al., 2014; Collins et al., 2018; Scharfenberger et al., 2019). All of these factors contribute to eDNA degradation and subsequently lower detection probabilities (Jo et al., 2019; Tsuji et al., 2017).
Species classed as cold water were detected in late autumn, mostly in tributary parts of the reservoirs. Specifically, coldwater salmonids were detected in running waters as expected due to higher water temperature at the surface and low DO concentrations in the colder, lower layers of standing water. eDNA accumulation in deep hypolimnetic water was not supported by our data. Species typically associated with deep open water, such as Coregonus spp., are naturally missing in artificial water bodies and where stocked, their presence is often restricted by lack of DO during summer stratification (Kottelat & Freyhof, 2007). The effects of DO and locality are further masked by tight correlations with temperature, trophic status and water age. For robust statistical analyses, more seasons and an equal number of habitats should be sampled. We selected two contrasting seasons and the number of habitats was derived by each reservoir´s morphology to achieve a reasonable compromise between investigator effort and information gained.
In our study, fish communities in canyon-shaped reservoirs were heavily driven by trophic status along the longitudinal gradient. The proxy of total ChlA concentration was positively correlated with temperature, and species associated with warmer water were detected in samples with higher trophic status. European eel was the only species strongly associated with trophic status independent of temperature. This species is an ambush predator, often with nocturnal activity (Trancart et al., 2019), and so may benefit from the dense water blooms and artificial darkness created by high trophic status. Community composition and distribution of particular species were found to change in response to trophic status in freshwater (Lawson Handley et al., 2018) and marine (Djurhuus et al., 2020) systems using eDNA metabarcoding. These findings and our own provide evidence that eDNA metabarcoding can be used to understand shifts in communities in response to environmental gradients.