A case study on joint species distribution modelling with bird atlas data: Revealing limits to species' niches

Growing interest in biodiversity mapping has spurred the development of species distribution atlases, often mainly based on citizen-science projects. Atlas data have been frequently exploited to model species' ecological niches and distributions on a species-by-species basis. However, spatial autocorrelation and phylogenetic relatedness among species complicate the statistical description of species' niches. Also, the effects of species' traits and co-occurrences on species-habitat relationship are commonly disregarded. In this work, we build a hierarchical multi-species model based on a major citizen-science project (the third Spanish breeding bird atlas) that simultaneously accounts for spatial, phylogenetic and trait-based dependencies. We predict the distributions of species niches, species richness and community traits along regional ecological gradients. Climate, habitat associations and species' traits all contribute (in this order) to structuring species' distributions. Species richness increases towards intermediate climatic conditions and with aquatic habitat cover and decreases with increasing forest and woody agricultural land cover. Species were distributed along regional climate gradients in accordance with their global thermal niches. Forest habitats favoured assemblages dominated by generalist, small-sized and cold-dwelling species with limited migratory behaviour. Increasing sampling effort augmented the model performance. Model performance was weaker for rare species and those with decreasing population sizes, likely due to their low niche saturation. Overall, we show that ecological relationships generalize from local to large scales and may be eludicated from atlases based on citizen-science mapping efforts.