Integrating species traits and habitat characteristics into models of butterfly diversity in a fragmented ecosystem

•We model environmental and species traits drivers of diversity in a fragmented area.•The 4th corner method identifies important environment–trait interactions.•We find that diversity patterns are primarily driven by environmental predictors.•Species traits and interactions may impact diversity, but more studies are required.•Our novel approach provides promising future directions for fragmentation research. The fragmentation of natural habitats has well-known impacts on biodiversity, although responses to fragmentation are often community and/or species-specific. One potential cause for this variation is that species traits may restrict the abilities of species to inhabit patches with certain characteristics (e.g., small amounts of available habitat). Although assessing the influence of both environmental attributes and species traits on diversity patterns across fragments is analytically challenging, it is crucial for better understanding the consequences of increasing anthropogenically driven land-use change. Using a multi-species, multi-site system, we explored both the causes of variation in butterfly diversity patterns in the fragmented landscape surrounding Prague, Czech Republic, and the utility of a novel modeling approach for better understanding the drivers of these patterns. Specifically, we integrated several environmental attributes and species traits predictors into a single modeling framework, explored variable selection through an unusual application of a secondary statistical technique, the fourth corner analysis, and then compared the fits of several models in order to determine whether diversity patterns were best explained by environmental attributes, species traits, or both. We found that in this system, butterfly diversity patterns are most influenced by habitat metrics, such as area and habitat heterogeneity. Species traits and environment–trait interactions had only minor impacts on diversity patterns. We suggest that this may be a result of the spatial scale and/or ecological context of this study system. The models, however, exhibited good fit to the data and showed strong predictive performance during model validation tests, suggesting that the techniques applied may provide useful tools for better understanding the effects of habitat fragmentation on natural communities.