Context-dependent differences in fish detection between eDNA metabarcoding and traditional methods

Environmental DNA (eDNA) metabarcoding is increasingly used to complement traditional fish surveys, yet its relative detection outcomes vary widely, and the factors associated with that variation remain unresolved. We conducted a global synthesis of 168 paired eDNA-traditional survey cases encompassing 3,841 fish species to quantitatively evaluate relative species-richness detection effectiveness at the community level and detection preference at the species level. At the community level, eDNA metabarcoding showed credibly greater relative species-richness detection than netting (95% CrI: 0.08–0.67). Higher relative species-richness detection by eDNA was supported for 12S-targeted assays (0.12–0.57) and temperate-zone surveys (0.09–0.67), whereas water volume, mean temperature and precipitation, and community species richness had limited linear influence. At the species level, detection preference was systematically associated with functional traits. In marine ecosystems, preferences for eDNA detection were supported for elongated (0.09–0.47), fusiform (0.03–0.38), and short-bodied fishes (0.01–0.44), whereas in freshwater they were supported for elongated (0.23–0.76), fusiform (0.06–0.60), and species with larger reported maximum body length (0.13–0.38). Habitat effects were ecosystem-dependent, favoring upper-water (0.37–0.99), mid-water (0.20–0.85), and deep-water (0.15–0.48) species in marine environments but lower-water (0.44–0.90) species in freshwater. These findings provide global empirical evidence for context-dependent differences in eDNA-based fish detection and a basis for standardizing its use in routine fish biodiversity assessment and ecosystem management.