Introduction
High quality water sources are essential for ecosystem service provisioning under global change, including climate and pollution (Baron et al., 2002). One effective way to maintain high quality water sources is water storage in reservoirs that have been created for drinking, agriculture and industrial purposes, flood protection, recreation, and hydropower generation among others (Labadie 2004). Despite the key importance of reservoirs, research has typically focused on natural water bodies whereas artificial or heavily modified water bodies are often neglected (Straškraba 2005). Deep reservoirs are prime, widespread examples of aquatic heterogeneous ecosystems with hydrological characteristics that are intermediate between lakes and rivers, and include lacustrine, transitional and riverine components (Wetzel 2001).
Deep temperate reservoirs are hydrologically complex environments where seasonal variation in temperature and flow conditions affects stratification and mixing patterns of the water column. River inflows along with an intricate outflow dam system further modify reservoir hydrodynamics and often result in currents and intrusion of water to deeper strata (Ford, 1990, Straškrabová et al., 1994). In summer, stable thermal stratification occurs with higher temperature (>20 °C) in the epilimnion and lower temperature (4 °C) associated with higher water density in the hypolimnion. Summer stratification is followed by autumn mixing of the entire water column, with gradual cooling to temperatures around or below 4 °C. During winter, the hypolimnion contains water of the highest density with temperature close to 4 °C, and surface water has lower density with possible ice coverage. The stability of the water column in winter depends on weather conditions (e.g. ice cover, wind, precipitation) and water level manipulations. In canyon-shaped reservoirs, spatial heterogeneity is also associated with longitudinal gradients (Thornton et al., 1990). Nutrient rich tributaries support phytoplankton production, which decreases towards the dam (Rychtecký & Znachor, 2011). A transition between the littoral and pelagic zones represents an additional source of reservoir spatial heterogeneity as the littoral zone differs in many aspects from more homogeneous pelagic habitats (Říha et al., 2015). Littoral habitats can be complex with macrophyte, rock and root configurations, or simple (e.g. sandy or muddy beaches), creating diverse environments affecting species composition, abundance and spatial distribution of the whole trophic web, from phyto- and zooplankton to fish (Prchalová et al., 2009a).
Fish have both direct and indirect effects on all trophic levels and are well-adapted to the various environmental conditions of heterogeneous temperate freshwaters. Therefore, they are often used as ecological indicators (EC, 2000). The long lifespan of many species also means that fish communities reflect changes over decadal timescales (Blabolil, et al., 2017b). Reliable assessment of fish community composition in reservoirs has typically required a combination of various sampling methods. However, established sampling methods are relatively laborious, expensive, sometimes destructive, taxonomically biased, and their application is case-specific, thus these limitations often result in poor accuracy and precision (Kubečka et al., 2009). In Europe, gillnetting is commonly used to assess fish populations, but this type of fish survey is extremely invasive (Blabolil, et al., 2017a). Therefore, efforts should be invested in the implementation of less-invasive and more universal monitoring strategies. Traditionally, visual census and hydroacoustics have been used for non-invasive surveys, but both have limitations. The accuracy of visual census depends on water clarity (Holubová et al., 2019), whereas hydroacoustics application is constrained by water depth and results do not provide species composition (Baran et al., 2017). Recently, environmental DNA (eDNA) metabarcoding has come to the fore as an innovative method for aquatic biodiversity biomonitoring (Taberlet et al., 2018).
eDNA refers to genetic material released by organisms into their environment, which can be captured and analyzed to enable non-invasive, highly sensitive species detection (Taberlet et al., 2018). Aquatic ecosystems are particularly suited to eDNA analysis as water is an effective medium for deposition and transport of intracellular or extracellular organismal DNA from aquatic, semi-aquatic and terrestrial organisms (Deiner et al., 2016; Harper et al., 2019b). All water bodies can be easily sampled to collect eDNA, including temporary pools (Bylemans et al., 2019), rivers (Deiner et al., 2016), ponds (Harper et al., 2019a), large heterogeneous lakes (Lawson Handley et al., 2019) and marine environments (Djurhuus et al., 2020). eDNA metabarcoding is an especially powerful and cost-efficient approach for biomonitoring as it allows whole communities to be screened (Hering et al., 2018). During metabarcoding, eDNA is PCR-amplified using broad-range primers and sequenced on a high-throughput sequencing platform (Deiner et al., 2017). However, species composition and distribution inferred from eDNA metabarcoding must account for environmental conditions that influence eDNA dynamics. Various factors affect eDNA persistence and degradation, such as temperature, ultraviolet (UV) light intensity, and oxygen concentration, that in turn determine microbial action and metabolic activity of aquatic organisms (Barnes et al., 2014). The spatial distribution of eDNA in a given water body can be influenced by species’ habitat preferences and activity (de Souza et al., 2016), oxygen and trophic conditions (Lawson Handley et al., 2019), or water movements caused by inflows and outflows, precipitation, wind, convective currents etc. (Jeunen et al., 2019). Background data characterizing environmental conditions and their possible links to eDNA dynamics are desperately needed, but are often not provided in published studies (Nicholson et al., 2020; Ruppert et al., 2019).
The aim of our study was to use eDNA metabarcoding to assess the fish communities in three Czech Republic reservoirs across two seasons. We compared species composition at multiple sites characterizing the spatial gradient in each of the reservoirs. We also tested the relationship between eDNA site occupancy and scores derived from conventional sampling for each taxon. Finally, we analyzed the relationship between species diversity and reservoir spatial heterogeneity.