Environmental DNA offers great potential as a biodiversity monitoring tool. Previous work has demonstrated that eDNA metabarcoding provides reliable information for lake fish monitoring, but important questions remain about temporal and spatial repeatability, which is critical for understanding the ecology of eDNA and developing effective sampling strategies. Here, we carried out comprehensive spatial sampling of England's largest lake, Windermere, during summer and winter to (1) examine repeatability of the method, (2) compare eDNA results with contemporary gill-net survey data, (3) test the hypothesis of greater spatial structure of eDNA in summer compared to winter due to differences in water mixing between seasons, and (4) compare the effectiveness of shore and offshore sampling for species detection. We find broad consistency between the results from three sampling events in terms of species detection and abundance, with eDNA detecting more species than established methods and being significantly correlated with rank abundance determined by long-term data. As predicted, spatial structure was much greater in the summer, reflecting less mixing of eDNA than in the winter. For example Arctic charr, a deepwater species, was only detected in deep, midlake samples in the summer, while littoral or benthic species such as minnow and stickleback were more frequently detected in shore samples. By contrast in winter, the eDNA of these species was more uniformly distributed. This has important implications for design of sampling campaigns, for example, deep-water species could be missed and littoral/benthic species overrepresented by focusing exclusively on shoreline samples collected in the summer. K E Y W O R D S eDNA, fish, lakes, metabarcoding, monitoring | 27 LAWSON HANDLEY Et AL.

1. Accurate, cost-effective monitoring of fish is required to assess the quality of lakes under the European Water Framework Directive. Recent studies have shown that environmental DNA (eDNA) metabarcoding is an effective and non-invasive method, which can provide semi-quantitative information about fish communities in large lakes.2. This study further investigated the potential of fish-based eDNA metabarcoding as a tool for lake assessment by collecting and analysing water samples from eight Welsh lakes and six meres in Cheshire, England, with well-described fish faunas.Water samples (N = 252) were assayed using two mitochondrial DNA regions (Cytb and 12S rRNA).3. eDNA sampling indicated the presence of very similar species in the lakes compared to those expected on the basis of existing and historical information. Firstly, 24 species were detected, with a total of 111 species occurrences in the lakes studied using eDNA. Secondly, there was a significant positive correlation between expected faunas and eDNA data in terms of confidence of species occurrence (Spearman's r = 0.74, df = 109, p < 0.001). Thirdly, eDNA data can estimate relative abundance with the standard five-level classification scale ('DAFOR').Lastly, four ecological fish communities were characterized using eDNA data which agree with the predefined lake types according to environmental characteristics. Synthesis and applications.There are some limitations when using conventional captured-based methods for surveying species richness and relative abundance, such as morphological identification bias, difficulties in recording small-bodied, rare and/or elusive species and destructive impacts on the environment. This study provides further evidence that environmental DNA metabarcoding outperforms other captured-based survey techniques in a wide range of lake types for community-level analysis whether in species detection, relative abundance estimate using the standard five-level classification scale or characterization ecological fish communities. Therefore, the fish-based environmental DNA metabarcoding, a non-invasive genetic method, has great potential as an assessment tool for lake quality under the European Water Framework Directive.