Mechanistic drivers of fish monitoring discrepancies between eDNA and traditional sampling

Environmental DNA (eDNA) metabarcoding is increasingly adopted for fish monitoring, yet its performance relative to traditional methods remains variable across methodological and environmental contexts. To resolve this uncertainty, we conducted a global synthesis of 168 paired eDNA-traditional survey cases encompassing 3,900 fish species to quantitatively evaluate community-level detection effectiveness and species-level detection preference. At the community level, eDNA metabarcoding significantly outperformed netting (effect size 95% credible interval: 0.08–0.67) and matched other approaches. Detection effectiveness was highest with 12S primers across ecosystems (0.12–0.56), whereas other single primers increased false negative risks. eDNA metabarcoding performed best in temperate (0.10–0.66) climatic zones, while water volume, local abiotic variables (temperature and precipitation), and community species richness showed limited linear influence. At the species level, detection preference was systematically driven by functional traits: eDNA metabarcoding preferentially detected elongated (0.11–0.29), fusiform (0.12–0.26), and larger-bodied (0.05–0.15) species globally. In contrast, habitat preference effects were ecosystem-dependent, favoring upper- and deep-water species in marine environments (Surface: 0.47–0.98; Mid: 0.16–0.76; Deep: 0.12–0.33) but bottom-water species in freshwater (Lower: 0.28–0.6; Deep: 0.12–0.48). These findings provide robust empirical evidence quantifying the mechanistic drivers of eDNA performance, supporting its standardized integration into routine biodiversity assessment and ecosystem management.