Integrating Dam-Induced Effects into a Bayesian eDNA-Hydrodynamic Model to Improve Fish Monitoring in Regulated Rivers

Abstract

Dam-induced hydrological changes threaten river biodiversity, underscoring the need for accurate monitoring. Environmental DNA (eDNA) offers a sensitive, noninvasive tool, but most studies overlook dam-driven flow and sediment alterations. Integrating eDNA with hydrological models that account for regulation is crucial to improve biodiversity assessments and guide conservation in modified rivers. Here, we developed a Bayesian framework that integrates prior knowledge of fish eDNA characteristics with hydrodynamic modeling and species distribution modeling, explicitly incorporating dam-induced effects on eDNA settling. Our analysis revealed that 98.8% of fish eDNA-adsorbing particles were ≤30 μm, exhibiting sedimentation velocities between 0.001 and 0.403 mm/s. Mean settling velocities differed markedly between reaches, averaging 0.041 in nondammed sections and 0.073 mm/s under dam influence. Incorporating decay rates and settling velocities into the Bayesian model improved the spatial accuracy of fish genus presence prediction. Relative to conventional hydrological models that do not incorporate dam effects, our approach yielded a 18.3–22.9% gain in predictive performance, with model outputs showing strong concordance with both observed fish distributions and environmental covariate responses. Overall, this study introduces a process-based framework addressing eDNA transport uncertainties in regulated rivers, incorporating hydraulic effects probabilistically, it improves ecological reliability of eDNA monitoring and supports targeted conservation under anthropogenic flow alteration.