The potential for environmental DNA (eDNA) to disperse widely from source organisms enables high detection efficiency but raises questions about eDNA's ability to differentiate fine-scale spatial patterns relative to conventional fish capture data.

We evaluate these questions in the St. Louis River estuary—a hydrologically and spatially complex coastal system within Lake Superior that supports a diverse assemblage of resident and migratory fish species—via comparison of eDNA metabarcoding (12S and 16S loci) to multigear capture survey data from 2 years and two seasons.

The eDNA and capture surveys collectively yielded 68 fish species: 2 species detected only by capture, 27 detected only by eDNA, and 39 shared across both survey types but having generally higher occurrence frequencies with eDNA than capture. Six species detected only by eDNA were unexpected, having no prior records in the Lake Superior basin. Data from paired eDNA and capture stations showed little relationship between the two survey types, with capture yielding species at stations that eDNA did not, eDNA detecting species in different habitats and distant locations from any captures, and assemblage patterns homogenized in eDNA surveys relative to capture surveys.

Our study finds that eDNA is a sensitive tool for assessing species presence at the system scale but that capture surveys may better yield the fine-scale spatial distribution information of interest to fisheries and habitat managers, especially in spatially and hydrologically complex systems.