Hidden pitfalls in aquatic eDNA sampling: A passive eDNA sampler reveals method-dependent biodiversity inference in freshwater systems
Abstract
Environmental DNA (eDNA) metabarcoding has become a powerful tool for freshwater biodiversity assessment, yet the physical architecture of eDNA samplers remains an underexamined source of methodological bias. Filtration is routinely treated as a neutral preparatory step; however, filter material, pore geometry, matrix structure, exposure duration and clogging dynamics each influence the particle-size fractions retained and, consequently, the community inferences derived. Here we introduce and evaluate the Environmental Passive Omnidirectional Device (ePOD), a low-cost passive sampler comprising a perforated high-density polyethylene housing fitted with a 1 µm nylon filtration matrix, benchmarked against conventional 0.45 µm polyethersulfone Sterivex active syringe filtration. Paired 24 hours ePOD deployments and 1 litre Sterivex filtrations were collected at three sites along the Ovens River, Victoria, Australia, and fish and vertebrate eDNA was profiled using mitochondrial 12S V5 amplicon sequencing, DADA2-based amplicon sequence variant (ASV) denoising and the MIDORI2 reference database for taxonomic assignment. Following control-based filtering and taxonomic curation, 6,131,765 reads were retained as fish-assigned, with a further 41,866,121 reads excluded as non-fish, unassigned, low-confidence or contamination-associated sequences. The ePOD yielded higher observed fish ASV richness (145 vs 138), Shannon diversity (3.82 vs 3.74) and Chao1 estimated richness (162 vs 154) than Sterivex filtration. Method-exclusive ASVs accounted for 16.6% of the ePOD fish dataset compared with 11.6% for Sterivex, and four conservation-relevant taxa, Macquaria australasica (Macquarie perch; EPBC Act 1999: Endangered), Maccullochella peelii (Murray cod; EPBC Act 1999: Vulnerable), Perca fluviatilis (European perch) and Anguilla reinhardtii (longfin eel), were detected exclusively by ePOD. The substantial non-target vertebrate signal in the broader 12S output reinforces the need for contamination-aware interpretation, robust positive and negative controls, and transparent bioinformatic curation. Collectively, these findings demonstrate that sampler architecture is a primary determinant of eDNA-based fish biodiversity inference, and they establish the ePOD as a scalable, infrastructure-independent passive sampling platform suited to freshwater monitoring programmes where active filtration is constrained by turbidity, field logistics or equipment availability.