A global consistent positive effect of urban green area size on bird richness

Abstract

Background: Although the species-urban green area relationship (SARu) has been analyzed worldwide, the global consistency of its parameters, such as the fit and the slope of models, remains unexplored. Moreover, the SARu can be explained by 20 different models. Therefore, our objective was to evaluate which models provide a better explanation of SARus and, focusing on the power model, to evaluate the global heterogeneity in its fit and slope. Methods: We tested the performance of multiple statistical models in accounting for the way in which species richness increases with area, and examined whether variability in model form was associated with various methodological and environmental factors. Focusing on the power model, we analyzed the global heterogeneity in the fit and slope of the models through a meta-analysis. Results: Among 20 analyzed models, the linear model provided the best fit to the most datasets, was the top ranked model according to our efficiency criterion, and was the top overall ranked model. The Kobayashi and power models were the second and third overall ranked models, respectively. The number of green areas and the minimum number of species within a green area were the only significant variables explaining the variation in model form and performance, accounting for less than 10% of the variation. Based on the power model, there was a consistent overall fit (r 2 = 0.50) and positive slope of 0.20 for the species richness increase with area worldwide. Conclusions: The good fit of the linear model to our SARu datasets contrasts with the non-linear SAR frequently found in true and non-urban habitat island systems; however, this finding may be a result of the small sample size of many SARu datasets. The overall power model slope of 0.20 suggests low levels of isolation among urban green patches, or alternatively that habitat specialist and area sensitive species have already been extirpated from urban green areas.