Extended Data Fig. 7: Predictors of dispersal limitation in birds.

Results shown are outputs of phylogenetic least squares model predicting dispersal limitation (nHWI) across all bird species sampled, including long-distance migrants (swallow image, dark bars; n = 1034); only resident species and short distance/partial migrants (thrush image, medium bars; n = 921); or resident species only (pitta image, pale bars; n = 858). Panels present three sets of models with increasing complexity: a univariate model with single predictor (a,d), and multivariate models with two (b,e) and three (c,f) predictors. Each predictor is calculated at the species level by averaging across landscapes where each species is present. Disturbance (red) is calculated as the proportion of species breeding range which overlaps areas of high natural (e.g. storms, glaciers, fires) or anthropogenic (e.g. forest loss) disturbance. Absolute latitude (yellow) is calculated as the centroid latitude of the species breeding range. Seasonality (blue) is calculated as the standard deviation of mean monthly temperature values throughout the year, averaged across all grid cells in the breeding range. a–c, Effect size estimates are given with 95% confidence intervals; a negative effect indicates reduced dispersal limitation (that is increased dispersal ability). R² and AIC values are calculated for full sample models only. d–f, Proportion of independent variation explained by each model covariate, calculated using hierarchical partitioning.