Abstract
Clarifying the effect of the sampling protocol on the detection of environmental DNA (eDNA) is essential for appropriately designing biodiversity research. However, technical issues influencing eDNA detection in the open ocean, which consists of water masses with varying environmental conditions, have not been thoroughly investigated. This study evaluated the sampling effort for the metabarcoding-based detection of fish eDNA using replicate sampling with filters of different pore sizes (0.22 and 0.45 μm) in the subtropical and subarctic northwestern Pacific Ocean and Arctic Chukchi Sea. The asymptotic analysis predicted that the accumulation curves for detected taxa did not saturate in most cases, indicating that our sampling effort (7 or 8 replicates, corresponding to 10.5-40 L of filtration in total) was insufficient to fully assess the species diversity in the open ocean and that tens of replicates or a substantial filtration volume were required. The Jaccard dissimilarities between filtration replicates were comparable with those between the filter types at any site. In subtropical and subarctic sites, turnover dominated the dissimilarity, suggesting that the filter pore size had a negligible effect. In contrast, nestedness dominated the dissimilarity in the Chukchi Sea, implying that the 0.22 μm filter could collect a broader range of eDNA than the 0.45 μm filter. Therefore, the effect of filter selection on the collection of fish eDNA likely varies depending on the region. These findings highlight the highly stochastic nature of fish eDNA collection in the open ocean and the difficulty of standardizing the sampling protocol across various water masses.