Modelling eDNA transport in river networks reveals highly resolved spatio-temporal patterns of freshwater biodiversity

Abstract The ever-increasing threats to riverine biodiversity call for the development of novel approaches for a complete assessment of biodiversity across highly resolved spatial, temporal and taxonomic scales. Past studies on riverine biodiversity patterns were often restricted to spatially scattered data, focused on specific taxonomic groups, and disregarded the temporal dimension, preventing a universal understanding of relationships between biodiversity and stream size across spatial, temporal and taxonomic scales. Recent advances in the joint use of environmental DNA (eDNA) data and novel mechanistic models for eDNA transport in river networks have the potential to uncover the full structure of riverine biodiversity at an unprecedented spatial resolution, hence providing fundamental insights into ecosystem processes and offering a basis for targeted conservation measures. Here, we applied a mechanistic model (i.e., the eDITH model) to a metabarcoding dataset covering three taxonomic groups (fish, invertebrates and bacteria) and three seasons (spring, summer and autumn) for a 740-km 2 Swiss catchment, sampled for eDNA at 73 sites. Using the mechanistic model, we upscaled eDNA-based biodiversity predictions to more than 1900 individual reaches, allowing an assessment of patterns of α - and β -diversity across seasons and taxonomic groups at a space-filling, fine scale over the whole network. We found that both predicted α - and β -diversity varied considerably depending on both season and taxonomic group. Predicted fish α -diversity increased in the downstream direction at all seasons, while invertebrate and bacteria α -diversity either decreased downstream or was not significantly related to position within network, depending on the season. Spatial β -diversity was mostly found to be decreasing in the downstream direction, and this was the case for all seasons for bacteria. Temporal β -diversity was mostly found to be increasing downstream. In general, genus richness values predicted by the model were found to be higher than those obtained by directly analyzing the eDNA data. Overall, stream size (subsumed by drainage area) was generally a poor predictor of patterns of predicted α - and β -diversities. Conversely, riverine biodiversity is shaped by a complex interplay of environmental variables, abiotic and biotic factors, which need be taken into account for a correct assessment of its structure.