Increased tolerance to humans among disturbed wildlife

Human disturbance drives the decline of many species, both directly and indirectly. Nonetheless, some species do particularly well around humans. One mechanism that may explain coexistence is the degree to which a species tolerates human disturbance. Here we provide a comprehensive meta-analysis of birds, mammals and lizards to investigate species tolerance of human disturbance and explore the drivers of this tolerance in birds. We find that, overall, disturbed populations of the three major taxa are more tolerant of human disturbance than less disturbed populations. The best predictors of the direction and magnitude of bird tolerance of human disturbance are the type of disturbed area (urbanized birds are more tolerant than rural or suburban populations) and body mass (large birds are more tolerant than small birds). By identifying specific features associated with tolerance, these results guide evidence-based conservation strategies to predict and manage the impacts of increasing human disturbance on birds.

The positive relationship between mean and standard deviation of FID is caused by an envelope constraint relationship that extends to the origin (i.e. there is no variation when mean FID is zero). Therefore, animals with larger mean FIDs also typically have larger standard deviations in FID. Because Hedges'g effect sizes were calculated as the mean differences between FIDs in low and high human disturbance divided by their pooled standard deviation 1 , the absence of interaction implies that our effect sizes estimates genuinely reflect the magnitude of mean FID differences (i.e. they were not biased by any potential difference in the variance of FIDs as a function of human disturbance level).

. Funnel plots of (a) all taxa and (b) birds-only meta-analyses
made with data from meta-analytic models in which study identities and phylogeny were used as random factors. If there were an obvious bias, it would be seen by relatively more points on the left side than on the right side in the bottom half of the funnel plots (i.e. more populations with large effect size than small effect size in studies with low sample sizes).
Relatively symmetrical distributions of effect sizes indicate that studies surveyed were not biased in their reporting of significant effects. This inference, based on visual assessment, was confirmed by the Egger's regression test, which found little evidence of publication bias (All taxa: intercept: -0.22, P = 0.243; Birds-only: intercept = -0.17, P = 0.397). Horizontal dashed line indicates zero effect size. Negative values illustrate tolerance of human disturbance. Taxa did not differ in their degree of tolerance of human disturbance (Q b = 1.21, df = 2, P = 0.54). There was substantial heterogeneity among effect sizes both in the meta-analysis including the three major taxa (I² total = 90.96%, I² between-study = 46.89%, I² species = 7.91%, I² withstudy(residuals) = 36.16 %) and in the birds-only meta-analysis (I² total = 88.99%, I² study = 40.84%, I² species = 9.59%, I² residual = 38.56%).   Birds species occurring in rural-urban habitat contrast had smaller mean body masses (Unequal variance t-test; t = 4.794, df = 202.531, P < 0.001) and larger variance (Levene's test; F 1,455 = 9.493, P < 0.001) than species from other habitats contrasts. These results indicate that the lower importance of clutch size in explaining bird's tolerance in the rural-urban habitat contrast was not caused by a reduction in clutch size variability in urban species, which could be caused by a reduction in their mean body size in urban places. Results are shown both from a meta-analysis using the full data set (all birds) and from a metaanalysis focusing on the contrast between rural and urban populations. Values are average coefficients of models (estimate) and their associated standard error (SE), and the importance of each factor in explaining species responses to human disturbance (the closer to 1, the most important the factor). Habitat contrasts presented as "low vs. high" mean contrast between populations with low and high human disturbance within a given habitat type. whereas values within white cells are from covariates used in model selection using only birds from rural-urban habitat contrasts. The highest correlations found were between body mass and foraging habit (r = 0.42) and between body mass and habitat openness (r = -0.42).

Phylogeny reconstruction
To obtain the phylogeny of birds species for our dataset, we used the time-calibrated avian phylogeny 11 available at http://birdtree.org/. Time-calibrated phylogenies were also available for mammals 12 and lizards 13 . We used the function prune.sample of the R package Picante 14 v. 1.6-2 to prune the tree so that it only included species present in our data sets.
We used a combined phylogeny of birds, mammals and lizards ( Supplementary Fig. 9) to test for differences in the overall effect sizes of these taxa while accounting for their shared evolutionary history. Because we do not have good estimates of basal branch lengths, we only used the topology of the overall tree created by joining all the three phylogenies together. For phylogenetic meta-analysis, we ultrametricized the tree using Grafen's method 15  noted that when phylogenetic signal is weak like in our case, phylogenetic analysis effectively reduces to non-phylogenetic analysis. This is because each data point (i.e. effect sizes) can be considered to be independent of phylogenetic relatedness.