Environmental DNA (eDNA) metabarcoding is revolutionizing the monitoring of aquatic biodiversity. The use of eDNA has the potential to enable non-invasive, cost-effective, time-efficient and high-sensitivity monitoring of fish assemblages. Although the capacity of eDNA metabarcoding to describe fish assemblages is recognised, research efforts are still needed to better assess the spatial and temporal variability of the eDNA signal and to ultimately design an optimal sampling strategy for eDNA monitoring. In this context, we sampled three different lakes (a dam reservoir, a shallow eutrophic lake and a deep oligotrophic lake) every 6 weeks for 1 year. We performed four types of sampling for each lake (integrative sampling of sub-surface water along transects on the left shore, the right shore and above the deepest zone, and point sampling in deeper layers near the lake bottom) to explore the spatial variability of the eDNA signal at the lake scale over a period of 1 year. A metabarcoding approach was applied to analyse the 92 eDNA samples in order to obtain fish species inventories which were compared with traditional fish monitoring methods (standardized gillnet samplings). Several species known to be present in these lakes were only detected by eDNA, confirming the higher sensitivity of this technique in comparison with gillnetting. The eDNA signal varied spatially, with shoreline samples being richer in species than the other samples. Furthermore, deep-water samplings appeared to be non-relevant for regularly mixed lakes, where the eDNA signal was homogeneously distributed. These results also demonstrate a clear temporal variability of the eDNA signal that seems to be related to species phenology, with most of the species detected in spring during the spawning period on shores, but also a peak of detection in winter for salmonid and coregonid species during their reproduction period. These results contribute to our understanding of the spatio-temporal distribution of eDNA in lakes and allow us to provide methodological recommendations regarding where and when to sample eDNA for fish monitoring in lakes.

Various factors can influence the concentration of eDNA found in the environment and blur the link between eDNA results and in situ fish abundances, biomasses or even occurrences. We studied the influence of one abiotic factor, the temperature, and one biotic factor, the fish size class, on the amount of fish eDNA detectable in water. To do so, we conducted two controlled experiments using tench (Tinca tinca, L) placed into outdoor mesocosms during 1 week. The quantification of fish eDNA abundances was performed using quantitative PCR after filtrating a large volume of water (30 L). The Bayesian hierarchical ANOVAs performed on qPCR results did not detect the effect of size class on eDNA amount, but fish eDNA concentration was shown to increase significantly when temperature rose by 6°C (no effect detected at 3°C). This suggests that fish assemblage quantification of two thermally contrasting sites through eDNA is not directly comparable.