Sample‐To‐Sample Variability Impacts eDNA Quantification, With Implications for Estimating Species Abundance

ABSTRACT Environmental DNA (eDNA) is increasingly used to infer species abundance, but uncertainty remains about how well individual water samples reflect system‐wide eDNA concentrations. Currently, high levels of variation in estimated eDNA concentration among samples, even under similar conditions, limit the utility of eDNA estimates of abundance, especially in conservation and management scenarios. To investigate the sources and magnitude of variation in eDNA estimates, we analyzed eDNA from replicate water samples taken from mesocosms housing Lithobates sylvaticus tadpoles. We partitioned the variance of observed eDNA concentrations between biological replicates (distinct water samples) and technical replicates. We further tested whether stochastic variation in extraction efficiency (e.g., silica‐column DNA binding and elution) introduced variance using internal control DNA. We then trialed modifications of sampling methods that could be applied to improve precision, including increased water volume, spatially pooled subsamples, and modified filtration. The observed variance in eDNA concentrations was substantial, ~75% of which was attributable to variation among replicate samples from the same mesocosm and < 5% related to technical replicates; extraction‐related variance was negligible. Sampling modifications, like increasing sample volume and taking combined samples of multiple scoops, improved precision, while others highlighted potential trade‐offs between precision and accuracy. Our results suggest that sample‐to‐sample variation, even from seemingly homogeneous, controlled environments, can be substantial even under highly controlled conditions. This inherent variability imposes limits on the precision of abundance estimates derived from eDNA and underscores the importance of replication and protocol optimization in study design.