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

C01. The animal’s body size and the primers chosen have direct influence on the eDNA survey; low DNA concentration [49] or the lack of amplification [16] or could explain its absence. Additionally, since most non-fish species are microscopic or from meiofauna,, there is a poten- tial for false negatives, i.e., the erroneous absence of a species. This bias can affect diversity esti- mates and, consequently, impact comparisons between traditional surveys and eDNA assessments [50]. Regarding the ichthyofauna, the eDNA survey identified nine species unre- ported to date, which corresponds to a ~6% increase in all species in the inventory of the study area. Nine major fish orders were detected across all samples, reinforcing previous reports for the TI. We identified abundant carnivorous species such as O. aureus, O. nilotocus, C. piquiti, C. kelberi, C. ocellaris and Plagioscion squamosissimus. In fact, three of the most abundant spe- cies captured by traditional monitoring during 2021 [51] were also among the most frequently detected at multiple stations using eDNA (P. squamosissimus and C. kelberi), supporting our findings. We also detected two native Crenicichla species, one of the richest groups in terms of spe- cies among the South American cichlids [52]. From the six taxa never previously reported in the TI reservoir, Hemiodus orthonops was detected at high frequencies at all sampling points. Some species were detected using traditional methods but not using eDNA, including Crenici-

chla haroldoi, Cichla monoculus, Metynnis maculatus, Moenkhausia sp., Astyanax fasciatus and Leporinus, for instance. Most of the species identified through traditional methods but not detected by eDNA have been reported only sporadically in local monitoring reports. This may suggest a seasonal reproductive or migratory behavior that makes them less consistently detectable [53]. Furthermore, the slow water flow, as those found for most of our sites accord- ing to hydrodynamic modeling, may hinder the transport and preservation of eDNA over extended periods, making species detection more challenging [54, 55]. This potential for sea- sonal presence could explain the absence of these species in the eDNA samples, despite their detection through traditional survey techniques. Even though it is tempting to look for an eco- logical explanation for the presence or lack of some species in certain locations, the absence of a species should not be solely attributed to its lack of occurrence since it can be a false negative. Several factors like species occupancy [56, 57], eDNA persistence and quantity [58, 59], primer bias [16] and DNA concentrations decrease [49] may influence the distribution pattern of eDNA. Following the same logic, the frequency of DNA fragments in the heatmap represents a rough estimation of the abundance of the species Also factor such PCR and/or errors might produce false positives due errors can occur due to random mutations during amplification or misreading during sequencing [60]. For these reasons, the eDNA abundance and/or absence

should be interpreted with caution as it may reflect body size of individual species [61] or drift [62], even for fishes [63]. When compared to the short-term traditional monitoring conducted during 2020–2021, eDNA successfully detected 60% of all fish taxa captured using traditional methods and more than 80% of all fish genera identified during the last 50 years (S6 Table). Several studies com- paring eDNA metabarcoding and established monitoring methods have been conducted in rivers [19, 64, 65] or estuaries [52]. Consistent with our findings, most of these studies have shown that eDNA methods exhibit similar performance compared to traditional methods. For example, Goutte et al. (2020) found that eDNA was more efficient than traditional surveys last- ing 3 to 6 years. In traditional community ecology, which relies on specimen counting, the distribution of organisms is usually clustered under the influence of environmental and biological factors such as physical barriers, nutrient availability and competition. However, eDNA sampling is not bound by the same constraints. Instead, it is influenced by drift, sedimentation, and degra- dation rates [5, 58, 61, 63] and, once released, eDNA tends to homogenize in the water flow, PLOS ONE eDNA metabarcoding reservoir PLOS ONE | https://doi.org/10.1371/journal.pone.0314210 January 24, 2025 15 / 22