Decision making and best practices for taxonomy-free eDNA metabarcoding in biomonitoring using Hill numbers

Abstract

Environmental DNA (eDNA) metabarcoding raises expectations for biomonitoring to cover organisms that have hitherto been neglected or excluded. To bypass current limitations in taxonomic assignments due to incomplete or erroneous reference data bases, taxonomic-free approaches are proposed for biomonitoring at the level of operational taxonomic unites (OTUs). However, this is challenging, because OTUs cannot be annotated and directly compared to classically derived data. The application of good stringency treatments to infer validity of OTUs and the clear understanding of the consequences to such treatments is thus especially relevant for biodiversity assessments. We investigated how common practices of stringency filtering affect diversity estimates based on Hill numbers derived from eDNA samples. We collected eDNA at 61 sites across a 740 km ² river catchment, reflecting a spatially realistic scenario in biomonitoring. After bioinformatic processing of the data, we studied how different stringency treatments affect conclusions with respect to biodiversity at the catchment and site levels. The applied stringency treatments were based on the consistent appearance of OTUs across filter replicates, a relative abundance cut-off and rarefaction. We detected large differences in diversity estimates when accounting for presence/absence only, such that the detected diversity at the catchment scale differed by an order of magnitude between the treatments. These differences disappeared between the stringency treatments with increasing weighting of the OTUs’ abundances. Our study demonstrated the usefulness of Hill numbers for comparisons between data sets with large differences in diversity, and suggests best practice for data stringency filtering for biomonitoring.