Animal populations often display coherent temporal fluctuations in their abundance, with far-ranging implications for species persistence and ecosystem stability. The key mechanisms driving spatial population synchrony include organismal dispersal, spatially correlated environmental dynamics (Moran effect), and concordant consumer-resource dynamics. Disentangling these mechanisms, however, is notoriously difficult in natural systems, and the extent to which the biotic environment (intensity and types of biotic interactions) mediates metapopulation dynamics remains a largely unanswered question. Here, we test the hypothesis that compositional differences among communities (i.e, beta-diversity), used as a proxy of the differences in biotic interactions experienced by separated populations, reduces population synchrony. Using an extensive dataset of fish population abundance time-series across Europe, we provide evidence that higher beta-diversity is associated with reduced spatial population synchrony within river networks, and demonstrate that these effects are independent from geographic separation, environmental dissimilarity, and Moran effects. Although beta-diversity is commonly shown to promote metacommunity stability by reducing spatial synchrony in aggregate community attributes (e.g., total biomass), our study indicates that compositional heterogeneity provides a previously overlooked spatial insurance effect that influences metapopulation dynamics by promoting asynchrony between populations separated in space. These findings illustrate how community assembly across different locations within river networks contributes to metapopulation stability and persistence of individual species, and further highlights the implications of the loss in beta-diversity over time via biotic homogenisation.

We gathered long-term (>10 years) fish populations time series for 61 basins across Europe – providing the most continuous and consistent data series – from the RivFishTIME database (50). Sampling occurred during low flows, and standard protocols were maintained through time. We included basins with at least eight sites and two species. In addition, only species occurring in more than 80% of sampling events were included for a total of 48 species.