Paul Savary , Cécile Tannier, Jean-Christophe Foltête, Marc Bourgeois, Gilles Vuidel, Aurélie Khimoun, Hervé Moal, Stéphane Garnier

Context and objectives
Although urbanization is a major driver of biodiversity erosion, it does not affect all species equally. The neutral genetic structure of populations in a given species is affected by both genetic drift and gene flow processes. In cities, the size of animal populations determines drift and can depend on multiple processes whereas gene flow essentially depends on the ability of species to disperse across urban areas. Considering this, we tested whether variations in dispersal constraints alone could explain the variability of neutral genetic patterns commonly observed in urban areas. Besides, we assessed how the spatial distribution of urban green spaces (UGS) and peri-urban forests acts on these patterns.
Methods
We simulated multi-generational genetic processes in virtual populations of animal species occupying either UGS or forest areas (both considered as a virtual species habitat) within and around 325 European cities. We used three dispersal cost scenarios determining the ability of species to cross the least favorable land cover types, while maintaining population sizes constant among scenarios. We then assessed genetic diversity and genetic differentiation patterns for each city and each habitat type across the three cost scenarios.
Results
Overall, as dispersal across the least favorable land cover types was more constrained, genetic diversity decreased and genetic differentiation increased. Across scenarios, the scale and strength of the relationship between genetic differentiation and dispersal cost-distances varied substantially, alike previously observed empirical genetic patterns. Forest areas contributed more to habitat connectivity than UGS, due to their larger area and mostly peri-urban location. Hence, population-level genetic diversity was higher in forests than in UGS and genetic differentiation was higher between UGS populations than between forest populations. However, interface habitat patches allowing individuals to move between different habitat types seemed to locally buffer these contrasts by promoting gene flow.
Discussion and conclusion
Our results showed that variations in spatial patterns of dispersal, and thus gene flow, could explain the variability of empirically observed genetic patterns in urban contexts. Besides, the largest habitat areas and biodiversity sources are likely to be found in areas surrounding city centers. This should encourage urban planners to pay attention to the areas promoting dispersal movements between urban habitats (e.g., UGS) and peri-urban habitats (e.g.. forests), rather than among urban habitats, when managing urban biodiversity.