How much is enough? Examining the sampling effort necessary to estimate mean eDNA concentrations in lentic systems

Abstract

The concentration of eDNA in an environment can provide important ecological information of relevance for management and conservation, but little research has explored optimizing sampling strategies to estimate mean eDNA concentrations in natural environments. Inter‐replicate eDNA concentrations often exhibit right‐skewed “clustered” or “clumped” distributions, likely due to the stochastic capture of large “aggregate” particles with high eDNA copy numbers. This has important potential implications for modeling the resulting sampling effort necessary to accurately quantify eDNA concentrations. In a previous study, 17–20 Brook Charr eDNA samples were collected from 28 lakes in Québec, Canada. We explored how variation in eDNA concentrations within a lake was affected by several habitat characteristics. We then conducted a power analysis to determine the sampling effort (“minimum n”) necessary to accurately quantify mean lake eDNA concentrations and, using simulations, explored how a bimodal distribution of eDNA particle copy count could affect inter‐replicate variability. The median sample size such that 90% of sample mean estimates were within 20% of the “true” mean was 12.5; a sample size of 20 was sufficient to quantify mean concentrations in 21/28 lakes. We found no evidence that temperature or lake size impacted sample variability. We also found that variance among replicates was non‐linearly related to mean lake eDNA concentration across years: variability was lowest at low and high concentrations and highest at intermediate concentrations. We hypothesize that this resulted from the stochastic capture of large “aggregate” particles at intermediate concentrations; at low concentrations, aggregates were likely rarely captured and at high concentrations may represent a consistent component of total eDNA. Simulations demonstrated that these patterns can emerge from some bimodal eDNA particle “size” distributions. Overall, we conclude that sampling efforts in many previous studies (notably including the authors' own) were potentially low, emphasizing the need to increase spatial replication in lentic surveys.

Metabarcoding is (usually) more cost effective than seining or qPCR for detecting tidewater gobies and other estuarine fishes

“…Knowing how many locations to sample and how far apart to space samples to adequately sample an estuary remains uncertain for eDNA sampling ( Yates et al, 2023 ). Although the results suggest a rule of five for tidewater gobies, this might not apply to large systems or low-density tidewater goby locations.…”

A Hierarchical Model for eDNA Fate and Transport Dynamics Accommodating Low Concentration Samples

“…The sampling process describes the variation in sample concentrations collected at a site as a function of the site concentration. This distribution is typically right-skewed due to eDNA clumping in the water column and stochastic collection of more rare, larger “aggregate” particles (Furlan et al, 2016; Yates et al, 2023); therefore, we use a lognormal distribution (Carraro et al, 2018; Espe et al, 2022) to describe this variation. Conditional on the eDNA concentration at site i , we assume that the concentration in each collected sample, , varies following: where σ samp is the sampling process standard deviation on the log scale, which could be a function of site covariates such as the distance from the release location.…”