Comparing local- and regional-scale estimations of the diversity of stream fish using eDNA metabarcoding and conventional observation methods

Abstract

1. We present a performance evaluation of environmental DNA (eDNA) metabarcoding with MiFish-U/E primers to investigate local and regional diversities of stream fish species to examine potential effectiveness, limits and future remedies of this technique in large-scale monitoring. We hypothesised that eDNA inferences are more consistent with fish assemblages observed upstream than downstream due to a directional flow of river water. 2. River water was sampled at 102 sites in 51 rivers around Lake Biwa in the central part of Honshu Island, Japan, within 10 person-days, and fish species compositions inferred from eDNA and existing observational data were compared. Observation sites were chosen from the observational data that were within a certain distance (buffer range) of a water-sampling site along a river trajectory. The hypothesis of the detection bias of eDNA towards upstream assemblage was tested by comparing results with all of the observational data, data from a higher elevation and data from a lower elevation. The Jaccard dissimilarity index was plotted between the observational data and the eDNA estimates against the buffer range; the buffer range with minimum dissimilarity was chosen. 3. When using existing observational data from within 6 km upstream of the eDNA sampling sites, the eDNA results were the most consistent with the observational data and inferred 86.4% of the species reported (38/44), as well as two additional species. eDNA results also showed patterns consistent with known upstream–downstream turnover of related species and biogeographical assemblage patterns of certain species. 4. Our 10-person-days survey using the metabarcoding technique enabled us to obtain as much regional fish diversity data including the hypothesised pattern of eDNA detection with an upstream bias as the accumulated observational data obtained through greater amounts of time, money and labour. The problems regarding false-positive/negative detection were suggested in our survey; however, these should be decreased or removed by modifying the sampling methods and experimental procedures in future works. Therefore, we concluded this new tool to enable monitoring that has never been implemented, such as cross-nation, and even whole-Earth monitoring with the data at yearly, seasonal or finer temporal scales.