ABSTRACTEnvironmental DNA (eDNA) metabarcoding surveys can support the acquisition of extensive biodiversity data to support ecosystem monitoring and conservation actions. However, the optimization of eDNA metabarcoding project design is essential to capture spatio‐temporal heterogeneity of eDNA signals and maximize diversity detection. In this study, we developed a system‐specific approach to detect fish communities in kelp forests, by analyzing fine‐scale spatio‐temporal patterns in eDNA signals at two sites along the South African coastline, as well as testing the effect of biological replication and pooling of replicates on species detection. At each site, samples were collected at two stations along the shoreline at two depth zones, and this was repeated at two time points (24 h apart). We detected 113 operational taxonomic units (OTUs) across 32 families, but fewer than 20% of OTUs could be assigned to species, indicating that barcode reference libraries need to be drastically improved. We detected significant differences in communities across small spatial scales (< 600 m) and time points, suggesting that to best capture a site's diversity patterns, samples should be collected at multiple points and times within at least 24 h. To detect ~80% of the fish community, including some low abundance species, a minimum of four samples appear sufficient. In addition, a higher number of OTUs (76 vs. 65) were found in individual replicates than in any of the pools. However, pooling samples prior to sequencing can still detect valuable broad‐scale biodiversity patterns for monitoring and can offset the decrease in data resolution with the benefit of accumulating comprehensive data from increased sampling efforts over time. As a pilot investigation into how best to maximize kelp forest‐associated fish communities, this study provides a basis for optimizing sampling design for coastal eDNA‐based surveys in southern Africa and strengthens the development of long‐term eDNA monitoring programs to better support conservation and management actions.
Read full abstract