Detection effectiveness and preference of environmental DNA metabarcoding in fish monitoring: a global comparison with traditional approaches

Environmental DNA (eDNA) metabarcoding has emerged as a powerful, non-destructive technology for fish monitoring. However, a comprehensive, global-scale quantification of how methodological, abiotic, and biotic factors influence its community-level detection effectiveness and species-level detection preference has been lacking. To address this, we conducted a systematic global synthesis based on 169 paired cases, encompassing 3,396 fish species from 79 orders, where eDNA metabarcoding and traditional approaches were applied concurrently. Using Bayesian generalized linear mixed models, we assessed how methodological protocols (e.g., traditional survey approaches compared, water volume, primer type), environmental conditions (e.g., ecosystem type, climate), and community attributes (e.g., species richness) shape community-level effectiveness, and how fish functional traits (e.g., body size, habitat preference) influence species-level detection preference. We found that eDNA metabarcoding outperformed traditional approaches in most cases (126/169) and for most taxonomic orders (61/79). Community-level analysis showed that netting-based approaches were consistently less effective than eDNA metabarcoding, while multiple traditional approaches combined showed comparable effectiveness. The effectiveness of eDNA metabarcoding slightly increased with water sample volume, eventually plateauing at larger volumes. The use of 12S primers or multi-primer combinations performed better than other single primer uses. eDNA metabarcoding was particularly effective in subtropical regions and estuaries, and the influences of temperature and precipitation were generally weak. Additionally, eDNA metabarcoding performance slightly declined with increasing species richness in marine and freshwater ecosystems. Species-level analysis showed that eDNA metabarcoding had a significant preference for surface, midwater, and larger-bodied species. Our findings provide the first global empirical evidence that eDNA metabarcoding can effectively complement or even replace traditional approaches in specific scenarios, advancing its standardized application in aquatic biodiversity monitoring and conservation. In this dataset and code, Troph denotes Trophic, water layer denotes habitat, and detection denotes PI. Partial adjustments have been made to the data processing and model fitting sections. Should this code be used for model analysis, subsequent plotting code requires corresponding object adjustments, as the plotting code only corresponds to models that have been fully analysed.