Artificial light at night (ALAN) is currently recognised as an important environmental disturbance that influences habitats, fitness and behaviour of numerous organisms. However, its effect on bird community distribution on a large spatial scale still remains unclear. Therefore, I decided to use a predictive approach to test an assumption that artificial nightlight, as one of 73 predictors, determines taxonomic, functional and phylogenetic levels of an avian community. In order to safeguard inference from any inconsistency, I used not one but four indices describing functional diversity, two measures showing phylogenetic species richness, and one reflecting taxonomic diversity. For all these measures of species communities I developed two sets of Random Forest models: one set included ALAN as an additional predictor, while the other did not. Following cross validation tests as well as an independent evaluation of models, I demonstrated that artificial night light improved the performance of predictive models. Taxonomic species richness decreased linearly along with increasing artificial luminescence. Moreover, functional diversity showed a unimodal relation to ALAN, which meant that most niches were occupied on a moderate level of artificial lighting. Finally, phylogenetic diversity was under the highest pressure of ALAN, because even a minimal amount of artificial night lighting radically reduced this measure of biodiversity. On the basis of predictive maps, I also found that models which did not include urbanisation processes showed high values of avian biodiversity in regions where in fact they were low. Thus, I conclude that ALAN as a human footprint can play a key role when analysing the distribution of bird communities on large spatial scales.
