Utility of eDNA and occupancy models for monitoring an endangered fish across diverse riverine habitats

Environmental DNA (eDNA) studies show great promise for non-invasive surveys of aquatic organisms, but should account for imperfect detection and the influences of biotic and abiotic conditions on detection. We evaluated an eDNA protocol for Roanoke logperch (RLP) Percina rex, an endangered fish of the eastern United States occupying habitats ranging from cold, clear creeks to warm, turbid rivers. We assayed water samples from streams presumed likely to be occupied or unoccupied by RLP based on previous fish surveys. Data were analyzed using multi-scale occupancy models that estimated occurrence and detection probability at the scales of sites, replicate water filters, and replicate PCR reactions, and environmental influences on these probabilities. We detected RLP eDNA at 11 of 12 sites in occupied streams and no sites in presumed-unoccupied streams. In best-supported models, detection was positively related to an index of fish density, whereas other environmental factors had no consistent effects. This approach had a higher detection rate and lower sensitivity to sampling conditions than traditional techniques like snorkeling and electrofishing, suggesting it could provide a powerful tool for assessing the distribution of this and other rare fishes that occur across a wide range of fluvial habitats.