Abstract
The analysis of environmental DNA (eDNA) has revolutionized biodiversity assessments in aquatic ecosystems, enabling non‐invasive monitoring of fish communities across diverse regions. However, the global comparability of these eDNA datasets remains ambiguous due to heterogeneous sampling protocols and bioinformatic workflows across studies, making it difficult to assess how robust and comparable the biodiversity patterns inferred from these datasets actually are. Here, we conducted a meta‐analysis of 58 riverine fish eDNA metabarcoding studies, covering 1818 sampling sites worldwide, to evaluate the robustness of eDNA‐derived biodiversity patterns. We found that species richness estimates and metrics of community structure derived under a common bioinformatic workflow were overall consistent with those of original analyses, despite the relatively high variability in bioinformatic analyses in the respective original studies. Contrastingly, congruence of species identity varied more extensively across datasets, mostly reflecting different completeness and regional relevance of reference databases. Restricting taxonomic assignment to basin‐specific species pools improved species identification accuracy, while datasets lacking publicly accessible or well‐curated reference data were more prone to mismatches. Year of sampling had a positive effect on taxonomic congruence, such that more recent studies showed increased robustness, also reflecting improved reference database coverage and enhanced species‐level identification over time and overall method congruence in more recent years. Overall, the suitability and potential of eDNA for global biodiversity monitoring is corroborating overall robust biodiversity estimates, irrespective of the bioinformatic approaches. Our study underlines the effectiveness and need for further harmonization of bioinformatic workflows and strengthened region‐specific reference databases for improved taxonomic resolution and comparability across studies.