A Systematic Review of Sources of Variability and Uncertainty in eDNA Data for Environmental Monitoring
Chloé Mathieu, Syrie M. Hermans, Gavin Lear, Thomas R. Buckley, Kevin C. Lee, Hannah L. Buckley
Frontiers in Ecology and Evolution
Environmental DNA (eDNA) is becoming a standard tool in environmental monitoring that aims to quantify spatiotemporal variation for the measurement and prediction of ecosystem change. eDNA surveys have complex workflows encompassing multiple decision-making steps in which uncertainties can accumulate due to field sampling design, molecular biology lab work, and bioinformatics analyses. We conducted a quantitative review of studies published prior to December 2017 (n = 431) that had sampled eDNA
... from a variety of ecosystems and that had explicitly accounted for variability and uncertainty associated with eDNA workflows, either in their study design (e.g., replication) or data analysis (e.g., statistically modeling the spatiotemporal variation). We recorded differences among research studies in their spatial and temporal study design, the detected scales of natural variation in the study taxa, and how researchers measured and addressed the multiple sources of variability and uncertainty associated with the eDNA workflow. We show that relatively few studies used eDNA to understand temporal variation in biodiversity compared to spatial variation, and fewer described how uncertainties were addressed. We recommend increasing the number of temporal studies and to account for both natural variation and sources of uncertainty, such as imperfect detection, when undertaking eDNA surveys. Of studies that quantified spatiotemporal variation, this review identified gaps in the scales over which researchers have observed these patterns. Increasing the number of long-term and broad-scale eDNA studies will improve understanding of how useful eDNA is at scales relevant for monitoring the effects of environmental changes such as climatic shifts or land use change. Even where sources of spatiotemporal variation and uncertainty were accounted for, the effort in quantifying this variation differed among the different steps in the eDNA process, from field, to laboratory and bioinformatics procedures, depending on the type of community studied (micro-vs. macro-organism communities). We recommend more consistent experimental and modeling methods, accounting for spatiotemporal variation, and uncertainty in eDNA collection, and analysis, and incorporation of prior knowledge of sources of variability via Bayesian modeling approaches to account for uncertainties such as imperfect detection, to generate robust diversity estimates and increase the comparability of eDNA datasets for environmental monitoring across space and time.