Insights into the representativeness of biodiversity assessment in large reservoir through eDNA metabarcoding

Results Hydrodynamic modeling shows a lentic environment and minimal water exchange in the TI Reservoir Interpolation of bathymetry from the Três Irmãos (TI) Reservoir showed that the average depth of the region was 18.3 meters. The TI Reservoir is a predominantly lentic environment in both rainy and dry seasons, with more intense currents upstream, next to the Nova Avanha- dava (NA) dam, and downstream near the TI dam. Hydrodynamic modeling results show that the water flows from the NA to the TI dam. The water velocity is higher near the main inflow into the reservoir at the NA dam (upstream in the reservoir) in both rainy and dry seasons (S2 Fig). Upstream, currents are more intense due to the topography of the channel, reaching up to 1.33 m/s and 0.56 m/s in the rainy and dry season, respectively. In the central region, the flow is very slow, with very low velocities (< 0.05 m/s) in rainy and dry seasons during all the modeled periods in the sections closest to the TI dam (S2 Fig). In the regions just upstream from the TI dam, currents are more intense than in the central region, but approximately 10% weaker than in the upstream region, presenting averages of 0.04 m/s in the rainy season and 0.02 m/s in the dry season. Water renewal in the reservoir compartments varies between rainy and dry seasons. During floods, water renews faster, peaking at 50 days and then increases exponentially. Compartment 1 (closest to NA dam) renews 50% of its water in less than five days during the rainy season but takes 10 days in dry periods. In compartments 2, 4, and 6, a similar

trend is observed for the first 50 days. After 70 days, sections 1, 2, and 4 have more than 95% of their water renewed during floods. Sections 1 and 2 in dry periods reach over 95% renewal in 90 days, while compartment 4 reaches about 80%. The furthest compartments (12, 13, and 14) only receive "new" water during the rainy season, renewing up to 10% after 50 days (S3 Fig). The results indicated that the water flow in the arms tends to stagnate, regardless of the time of year, meaning there is a low exchange of water between the main flow and the arms. Taken together, the reservoir depth, water velocity, and water renewal rate supported the dis- tribution of stations for water sampling, where they are allocated to capture eDNA in the dif- ferent hydrological profiles of the reservoir and the arms. Bottom sampling and three-marker analyses improve the chances of fish eDNA capture The pilot sampling obtained an average of 1.5 million of 150 bp paired-end reads per depth. The raw sequence data are publicly available at the SRA database under the BioProject acces- sion code PRJNA1105563. After the read cleaning, amplicon merging, chimera removal, and clustering of identical amplicons, 615, 2,993, and 4,404 ASVs remained for 12S, 16S, and COI datasets, respectively (S3 Table). After applying the Bayesian classifier from QIIME2 and removing all off-target species, COI exclusively identified 24 taxa, while 16S found 3 and 12s, none. These datasets merged resulted in 27 non-fish taxa, which belonged to Rotifera, Cope- poda, Annelida, Algae and Viridiplantae clades (S4 Table).

Regarding ichthyofauna, the GLM revealed that sampling at the benthic bottom (25 m) increases the chance of capturing fish eDNA (S4 Fig). COI, 12S and 16S detected five, four and one taxa, respectively, resulting in 10 taxa. Four species were identified by both 16S and COI, one by 12S and COI, and one by 12S and 16S. Only 4 species were detected by all three mini- barcodes (S4 Table). PLOS ONE eDNA metabarcoding reservoir PLOS ONE | https://doi.org/10.1371/journal.pone.0314210 January 24, 2025 7 / 22