Landscape variables impact the structure and composition of butterfly assemblages along an urbanization gradient

How urbanization affects the distribution patterns of butterflies is still poorly known. Here we investigated the structure and composition of butterfly assemblages along an urbanization gradient within the most urbanized and densely populated region in France (Île-de-France). Using a method issued from artificial neural networks, i.e. self-organizing maps (SOMs), we showed the existence of four typical assemblages ranging from urban-tolerant species to urban-avoider species. We identified indicator species of these assemblages: the peacock butterfly (Inachis io) in urbanized areas, the swallowtail (Papilio machaon) in sites with intermediate human pressure, or the meadow brown (Maniola jurtina), the small heath (Coenonympha pamphilus) and the gatekeeper (Pyronia tithonus) in meadows around Paris. A discriminant analysis showed that the four assemblages were mainly segregated by landscape elements, both by structural variables (habitat type, proportion of rural areas and artificial urban areas, patch surface) and functional variables (distance to the nearest wood, artificial area and park). Artificial neural networks and SOMs coupled stepwise discriminant analysis proved to be promising tools that should be added to the toolbox of community and spatial ecologists.