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

where only C01 appears different from all others, and the segregation among the remaining locations becomes indistinguishable in the ordination space (Fig 5B). When analyzing the sums of contributions from each station to the six principal components, stations C01, C10 and C11 exhibited the highest indices (Fig 5C and S5 Fig). By using six PCs according to the alpha score results, which reached 98.7% of the total variation, the DAPC identified five dis- tinct groups: three groups comprised of one sample each from station C01, one group com- prised of one sample each from C06 and C11, and the fifth group comprised of all the remaining samples (Fig 5D). In addition, a single sample from station C15 presented equal chances of belonging to two distinct groups. The composition graph of all five clusters revealed that samples from C01 belong to differ- ent clusters, suggesting that even being sampled within the same station, the three obtained samples were different enough to be considered distinct community groups (Fig 5D). After these findings, using the samples with the highest contribution for the six PCs, rarefaction curves were computed for the datasets: i) containing all samples; ii) without C01; iii) without C11; iv) without C10; and v) containing only C01, C11, and C10 samples. For the complete dataset, as well as for those lacking one sample, the curves reached the plateau between 10–12 samples. The graphs indicated minimal changes in the plateau area, with slightly more notice- able differences observed when removing C01, suggesting that additional samples would be

needed to fully capture the observed diversity (Fig 6). When C10 and C11 were removed, the results were relatively similar and did not exhibit a clear alteration in the curve shape. The rar- efaction curve built using only C01, C10, and C11 samples almost reached the plateau area after nine samples, the total number (S5 Fig). Discussion This study provides an important insight into the optimal sampling strategy when conducting eDNA surveys in hydroelectric power plant reservoirs. Here, through the combination of hydrodynamic modeling and pilot sampling, we could delineate an optimal experimental design for the TI reservoir. In this study, we managed to perform a biodiversity baseline of a large reservoir for the first time in South America. Metabarcoding successfully detected a substantial number of species previously docu- mented in the literature over the past 50 years, along with nine fish species never reported in the reservoir before. We found that larger sampling sizes might not be the best approach when assessing biodiversity in lentic reservoirs using eDNA. Instead, it appears preferable to sample fewer locations with more replicates. Moreover, a one-year interval between sample collections did not seem to significantly influence diversity measures or the detection of species. Lastly, our data provides a valuable contribution to future biodiversity monitoring studies in reser- voirs, aiming to achieve optimal utilization of resources with lower costs and better cost-effec- tiveness through molecular data.

This study contributed 20 new previously unreported non-fish species for the TI area (~5%). Overall, there was a high diversity of non-fish taxa, with annelids and rotifers found across several samples (Fig 2). Copepods, on the other hand, were abundant but less diverse, mainly represented by the genera Thermocyclops and Mesocyclops. Chordates were predomi- nantly represented by birds with marine habitats, such as herons (Ardea cinerea) and Anatidae species (ducks, geese, and swans), found just upstream from the dam (C01-C03) and in the middle of the reservoir (C04, C07, C10, and C12). The only non-avian chordate detected was the capybara (Hydrochoerus hydrochaeris) in C01, a mammal commonly found in riverside and lakeside areas [47]. Contrary to what would be expected, the invasive golden mussel spe- cies Limnoperna fortunei, highly abundant in the TI HPP intake [48], was detected only in PLOS ONE eDNA metabarcoding reservoir PLOS ONE | https://doi.org/10.1371/journal.pone.0314210 January 24, 2025 13 / 22