Detection differences between eDNA and mid-water trawls are driven by fish biomass and habitat preferences

Marine scientific trawl surveys are commonly used to assess the distribution and population size of fisheries-related species, yet the method is effort-intensive and can be environmentally destructive. Sequencing environmental DNA (eDNA) from water samples can reveal the presence of organisms in a community without capturing them; however, we expect the detectability of taxa to differ between eDNA and trawl surveys, and understanding how species traits and population variables contribute to detection differences can help calibrate our expectations from each form of sampling. Here, we coupled eDNA metabarcoding and capture trawl surveys in British Columbia, Canada, to examine species traits that explain recurrent differences in detectability between the two methods, including habitat, body size, and biomass. At the regional scale, 17 of 23 fish species (74%) captured by the trawl were detected by eDNA metabarcoding, and 39 additional species were detected by eDNA sampling only. We found that eDNA metabarcoding disproportionately detected trawl-caught species with greater local biomass (i.e., greater biomass in the adjacent trawl). Fish detected only in eDNA had a greater range of body lengths and a broader range of habitat preferences outside the trawls' target size and sampling areas. Our results suggest that with our level of sampling, eDNA metabarcoding can adequately recapitulate detection of fish communities detected by trawl surveys, but with a bias toward fish of high population biomass and greater inclusion of fish from outside the trawled area.