AUTHOR=Ivanova Petya , Zlateva Ivelina , Dzhembekova Nina , Popov Ivan , Raykov Violin , Dimitrov Dimitar , Mihova Svetlana , Raev Yordan , Stefanova Kremena 

TITLE=A study of feasibility and detection sensitivity of environmental DNA for fish biodiversity monitoring and stock assessment in the Black Sea

JOURNAL=Frontiers in Marine Science

VOLUME=Volume 12 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2025.1648741

DOI=10.3389/fmars.2025.1648741

ISSN=2296-7745

ABSTRACT=IntroductionHuman activities and their effects on ecosystems are threatening biodiversity and causing a decline in species, population, and genetic diversity. Environmental DNA (eDNA) metabarcoding represents a powerful instrument to assess marine biodiversity, enabling an efficient, cost-effective, and non-invasive approach. eDNA metabarcoding technology has significantly progressed in recent years and has been widely employed to assess marine fish populations.MethodsThis study assessed its potential, detection sensitivity, and effectiveness for an initial evaluation of fish biodiversity in the Black Sea. The efficacy of eDNA was evaluated by contrasting metabarcoding results with trawling data obtained from two sampling campaigns (summer and autumn 2022) across 16 selected locations along the Bulgarian Black Sea coast. Mitochondrial genes 12S and 16S were utilized as standard markers for metabarcoding given their high specificity and sensitivity in amplifying fish DNA, and the availability of large reference databases. A multi-model analytical approach utilizing Bayesian regression and Generalized Additive Models (GAMs) was implemented to assess the relationships between environmental DNA (eDNA) 12S metabarcoding data and abundance indices obtained from catch-per-unit-effort (CPUE). The method addresses significant challenges in predicting abundance from eDNA, including scarce, noisy datasets, excessive zeros, and nonlinear predictor-response relationships. Both frameworks reliably detected biologically meaningful associations between eDNA signal strength, environmental gradients, and trawl-derived abundance.ResultsMetabarcoding, employing the 12S mitochondrial gene with MiFish-U primers, exhibited superior sensitivity in identifying a broader array of fish species, highlighting its efficacy in detecting eDNA traces of rare and migratory species. During the autumn survey, environmental DNA (eDNA) analysis identified 23 fish species, compared to 15 species recorded through trawl sampling. In the summer expedition, 12 species were detected using molecular methods, while trawl surveys recorded 9 species. The Bayesian framework provided robust uncertainty quantification and reliable inference with few samples, whereas the GAM framework effectively captured nonlinear environmental and spatial effects. The results indicate that eDNA data can replicate patterns observed in conventional trawl surveys. This supports the assumption that eDNA data may serve as a low-impact, scalable method for monitoring fish populations. Model uncertainty informs the need for more sampling or a redesign of covariates to improve the accuracy of eDNA-based abundance predictions.DiscussionOur findings demonstrate that the eDNA 12S primer combination exhibited superior taxonomic diversity compared to trawling, while necessitating reduced sampling effort. Furthermore, the eDNA-based approach can identify species across diverse life stages and sizes, unlike traditional trawling, which predominantly captures adult specimens within particular size categories. Although eDNA metabarcoding has limits in providing estimates for absolute biomass and size distributions, it provides significant insights on the presence of numerous fish species within the studied ecosystem. The consistency of results from multiple statistical frameworks supports the validity of eDNA counts as an abundance measure, provided that analytical models effectively account for dispersion and environmental variability. The Bayesian regression approach enabled thorough uncertainty quantification and dependable parameter inference with constrained sample sizes, whereas Generalized Additive Models (GAMs) provided an advanced representation of potentially nonlinear relationships among environmental DNA, environmental variables, and abundance.