Optimizing environmental DNA sampling effort for fish inventories in tropical streams and rivers

Environmental DNA (eDNA) metabarcoding is a promising tool to estimate aquatic biodiversity. It is based on the capture of DNA from a water sample. The sampled water volume, a crucial aspect for efficient species detection, has been empirically variable (ranging from few centiliters to tens of liters). This results in a high variability of sampling effort across studies, making comparisons difficult and raising uncertainties about the completeness of eDNA inventories. Our aim was to determine the sampling effort (filtered water volume) needed to get optimal inventories of fish assemblages in species-rich tropical streams and rivers using eDNA. Ten DNA replicates were collected in six Guianese sites (3 streams and 3 rivers), resulting in sampling efforts ranging from 17 to 340 liters of water. We show that sampling 34 liters of water detected more than 64% of the expected fish fauna and permitted to distinguish the fauna between sites and between ecosystem types (stream versus rivers). Above 68 liters, the number of detected species per site increased slightly, with a detection rate higher than 71%. Increasing sampling effort up to 340 liters provided little additional information, testifying that filtering 34 to 68 liters is sufficient to inventory most of the fauna in highly diverse tropical aquatic ecosystems.