Continuous eDNA Monitoring Reveals Rare Fish Diversity and Enables Filtering of Exogenous DNA Signals in a Megariver.

Megarivers support high biodiversity but challenge monitoring efforts, as conventional low-frequency environmental DNA (eDNA) surveys often miss rare taxa and may misclassify transient exogenous DNA. Here, we conducted daily eDNA monitoring for 1 year at a fixed site on the Yangtze River to evaluate the effects of sampling frequency and distinguish native from exogenous signals. Among 210 detected fish taxa, 124 were verified as native. Fish diversity exhibited strong temporal variation, peaking in warm seasons. Downsampling analyses revealed that quarterly sampling recovered only ∼30% of annual diversity─primarily missing rare taxa─whereas sampling every ≤3 days recovered >90%. To address exogenous contamination, we trained a Random Forest classifier using daily eDNA profiles. An "environmental coupling score", which quantifies synchrony between eDNA and water-quality dynamics, emerged as the strongest predictor. Achieving an AUC of 0.727, this model substantially outperformed conventional abundance-based thresholding. Together, our results demonstrate that continuous daily eDNA monitoring substantially improves the detection of rare species in hydrodynamically complex rivers. This study offers quantitative guidance on sampling design and provides a robust machine-learning framework to reduce misclassification of exogenous signals in large-river biomonitoring.