Validation of graph-based connectivity models using genetic data.

International audienceModelling the functional connectivity of habitats is crucial for biodiversity conservation. By modelling potential dispersal paths among habitat patches, landscape graphs are often used to quantify landscape connectivity. While this approach seems promising, it often lacks biological validation. To ensure its ecological relevance, we assessed the ability of connectivity metrics calculated from landscape graphs to predict population genetic structure that closely reflects the dispersal of individuals, and thus functional connectivity. We modelled the habitat network of a forest bird species (Plumbeous warbler, Setophaga plumbea) in Guadeloupe using three graphs constructed either from expert opinion, habitat specialization indices, or a species distribution model (SDM). Genetic data (microsatellites) were also collected on 712 individuals in 27 populations. This genetic dataset was used as an empirical validation tool for the three landscape graphs, using two approaches: i) relating the cost-distances obtained from the graphs to between-population genetic distances, ii) relating the connectivity metrics obtained from the graphs to within-population genetic diversity. The large proportion of variance in genetic distances explained by least-cost paths and the strong correlation between connectivity metrics and genetic diversity indices demonstrate the ability of landscape graphs to model the influence of landscape connectivity on dispersal. In addition, our results provide insight into the relationships between construction costs and ecological relevance of landscape graphs, as the most complex approach (SDM) is not always the most efficient