Analysis of species interaction networks in a fish and amphibian floodplain metacommunity using eDNA metabarcoding.

Large floodplain rivers are among the most species-rich and complex systems, characterized by high spatiotemporal dynamics. The exchange between communities of different patches in space and in time depends on hydrological conditions that impact the distribution of species and their interactions. This complexity makes it particularly challenging to identify key features of communities, including species interactions, which are also influenced by dynamic dispersal patterns. We collected eDNA data for fish and amphibians across two floodplain systems along the Danube over three years, capturing a range of hydrological conditions from post-flood to extended dry periods. Using an approach based on Bayesian networks we analyze for species co-occurrence patterns accounting for spatial, temporal, and dynamic autocorrelation as well as environmental conditions. In the second step, we applied a graph-theoretic approach to depict and analyze the relationships between species and define discrete communities. Our results reveal that species differ in their migration intensity, as reflected in the varying significance of temporal and spatial autocorrelation within the system, i.e., having continuous impact in local communities or impact changes over time or hydrological conditions. Further, we identified different communities in the system, including one clearly delineated consisting of amphibians and a few stagnotopic fish species showing negative interaction with other fish communities, alongside more open fish communities, i.e., often showing positive interactions with other communities. Different species interactions such as predator-prey interactions within fish as well as between fish and amphibians, are well delineated in the network. Several invasive fish species are also strongly interacting, they show relatively high connectivity within the species network. Overall, our approach contributes to a more mechanistic understanding of species interactions in complex, dynamic systems.

This research acknowledged support from the Austrian Science Fund (FWF) project RIMECO (I 5006), the EU Projects H2020 MERLIN (grant agreement No 101036337), HEU Danube4all (grant agreement No 101093985) i-CONN’ H 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement number 859937. Furthermore, the Austrian Federal Ministry for Digital and Economic Affairs and the Christian Doppler Research Association supported the work via the Christian Doppler Laboratory for Meta Ecosystem Dynamics in Riverine Landscapes (CD Laboratory MERI).