LAUSR

Abstract Cities are human dominated ecosystems providing novel conditions for organisms. Research on urban biodiversity is rapidly increasing, yet it is still hampered by the partial spatial coverage of cities and because of existing taxonomic biases. Predictive models have proved to be a key tool to solve this shortfall. However, predictive models have rarely been used in urban ecosystems due to either the lack of sufficient species records or high-quality predictors (e.g. meaningful ecological maps). Here, we assemble a large cross-taxa inventory of 1446 species from 12 taxonomic groups, including several understudied invertebrate groups, sampled in 251 sites in Zurich, Switzerland. We investigate the species diversity distributions and the structure of species assemblages along artificial urban ecological gradients by applying predictive models. We find that the general species diversity distribution law, where assemblages are dominated by a few very abundant and frequent species, applied consistently across all taxonomic groups (3% of the species accounting for approximately 50% of abundance). Furthermore, only species of intermediate abundance and frequency are spatially structured along urban intensity gradients, with rare species numbers keeping constant even in the most urbanised parts of the city. In addition, we show that green areas with low mowing regimes are associated with higher species diversity in the majority of taxonomic groups. Hence, this suggests management relaxation as a low-cost solution to promote species richness. Our study demonstrates the potential of predictive modelling for addressing ecological questions in urban environments and to inform management and planning.