Genetic-based inference of densities, effective and census sizes of expanding riverine meta-populations of an invasive large-bodied freshwater fish ( Silurus glanis L. )

Effective (N _e ) and census (N _c ) population sizes are key eco-evolutionary parameters. Jointly estimating them have an important practical value for efficient conservation and wildlife monitoring and management. Assessing N _e and N _c remains however challenging for elusive, rare species or species inhabiting in complex habitats like large rivers. Genetic-based N _e estimations could help resolve complex situations, as only a handful of genotyped individuals are needed to estimate N _e , and then N _C can be subsequently using an N _e /N _C ratio. However, most N _e estimation methods are based on restrictive assumptions (e.g. Wright-Fisher model) making them inappropriate for inferring N _e and N _c for populations and species exhibiting complex dynamics. Here, we aimed at estimating N _e , N _C and densities for meta-populations of a large invasive freshwater fish (the European catfish Silurus glanis ) that has been introduced in the Garonne-Dordogne river basin (Southwestern France), using a framework that combines multiple data sources and approaches. First, we characterized spatial patterns of genetic variation using microsatellite genotype data, revealing a significant isolation by distance pattern informing about the species’ dispersal capacities. We then detected four genetically-distinct clusters of individuals coexisting in the river basin that might be the result of multiple introductions from different genetic sources. Further, we characterized the demographic expansion of the species at the river basin scale by analyzing data from a multidecadal demographic monitoring survey, and estimated a specific Ne/Nc ratio for this species. We finally combined all the gathered information to design four competing demo-genetic models accounting for all the complexity of S. glanis meta-populations inhabiting the river basin. We simulated data under these models and then inferred Ne, Nc and densities through approximate Bayesian computation and random forest procedures. We show how multiple genetic and non-genetic approaches can be combined to estimate N _e and N _c in hard-to-monitor meta-populations exhibiting complex demo-evolutionary dynamics.