The world\'s largest rivers are home to diverse, endemic, and threatened fish species. However, their sheer sizes make large-scale biomonitoring challenging. While environmental DNA (eDNA) metabarcoding has become an established monitoring approach in smaller freshwater ecosystems, its suitability for large rivers may be challenged by the sheer extent of their cross sections (>1 km wide and tens of meters deep). Here, we sampled fish eDNA from multiple vertical layers and horizontal locations from two cross sections of the lower reach of the Yangtze River in China. Over half of the ASVs (amplicon sequence variants) were detected in only a single combination of the vertical layers and horizontal locations, with ∼7% across all combinations. We estimated the need to sample >100 L of water across the cross-sectional profiles to achieve ASV richness saturation, which translates to ∼60 L of water at the species level. No consistent pattern emerged for prioritizing certain depth and horizontal samples, yet we underline the importance of sampling and integrating different layers and locations simultaneously. Our study highlights the significance of spatially stratified sampling and sampling volumes when using eDNA approaches. Specifically, we developed and tested a scalable and broadly applicable strategy that advances the monitoring and conservation of large rivers.

Environmental DNA (eDNA) technology has revolutionized biomonitoring in recent years; however, eDNA collection from aquatic environments generally relies on the time-consuming and equipment-dependent process of water filtration. Passive eDNA sampling deploys sorbent materials to capture eDNA from water, circumventing many problems associated with active filtration; yet, very few candidate materials have been systematically evaluated for this purpose. Here, we evaluated the ability of 12 different types of common loose sorbents and filter membranes to capture eDNA in laboratory and field experiments compared with conventional water filtration. Glass fiber filters (GF) outperformed all other materials in laboratory experiments with respect to their quantitative capacity to recover amphibian eDNA, with the eDNA yield increasing linearly with submersion time up to 72 h. Furthermore, GF rapidly (within 0.5 h) captured the eDNA of up to 71% of the total fish species in a lake, in addition to detecting the entire fish community by 8 h, as assessed by metabarcoding analysis. Our results demonstrate that GF could passively capture aqueous eDNA with a similar or greater efficiency than conventional methods, thus paving the way for convenient, effective, and eco-friendly eDNA sampling in aquatic environments.