Social information modifies the associations between forest fragmentation and the abundance of a passerine bird

Habitat loss and fragmentation are the main factors driving the occurrence and abundance of species in the landscape. However, the local occurrence and abundance of species may also depend on conspecific and heterospecific social information e.g. clues of animals’ presence or their voices. We investigated the impact of the interaction between different types of social information and forest fragmentation on the abundance of the song thrush, Turdus philomelos, in Central Europe. Three types of social information (attractive, repulsive, and mixed) and procedural control were broadcasted via loudspeakers in 150 forest patches that varied in size and isolation metrics. Repulsive social information (cues of presence of predator) decreased abundance of song thrush. Also, the repulsive social information changed the association between forest patch isolation, size and the abundance. Attractive social information (songs of the studied thrush) had no effect on song thrush abundance. However, the attractive social information reversed the positive correlation between habitat patch size and the abundance. Mixed social information (both repulsive and attractive) had no impact on the abundance nor interacted with habitat fragmentation. The observed effects mostly did not last to the next breeding season. Overall, our findings indicate that lands of fear and social attraction could modify the effect of habitat fragmentation on the species abundance but these effects probably are not long-lasting.


Figure S2 .
Figure S2.Effect of the mean number of other bird species (excluding song thrush) on the relative abundance index of song thrushes (differences in abundance index between 2018 and 2017, and between 2019 and 2017).Estimates from generalized additive mixed models with Gaussian error distribution.In every model total survey duration (in minutes) was included as the offset variable.Ribbons are 95 % confidence intervals.Explanations: * -p < 0.05, *** -p < 0.001.

Figure S3 .
Figure S3.Effect of temperature (A) and mean temperature (B) on the predicted abundance index of song thrushes in 2017 and relative abundance index (2019-2017), respectively.Estimates from generalized additive mixed models with Poisson (A) and Gaussian (B) error distribution.In every model survey duration (in minutes) was included as the offset variable.Ribbons are 95 % confidence intervals.Explanations: ** -p < 0.01, *** -p < 0.001.

Figure S4 .
Figure S4.Effect of mean cloudiness on the relative abundance index of song thrush (difference in abundance index between 2019 and 2017).Estimates from the generalized additive mixed model with Gaussian error distribution.Total survey duration (in minutes) was included as the offset variable.Ribbons are 95 % confidence intervals.Explanations: * -p < 0.05.

Figure S5 .
Figure S5.Effect of the starting survey time (minutes from sunrise) on the abundance index of song thrushes each year.Estimates from generalized additive mixed models with Poisson error distribution.In every model survey duration (in minutes) was included as the offset variable.Ribbons are 95 % confidence intervals.Explanations: * -p < 0.05, *** -p < 0.001.

Figure S6 .
Figure S6.The effect of survey date (expressed as the number of days from 1st April) on the abundance index of song thrush.Estimates from generalized additive mixed models with Poisson error distribution.In every model survey duration (in minutes) was included as the offset variable.Ribbons are 95 % confidence intervals.Explanations: * -p < 0.05, *** -p < 0.001.

Figure S7 .
Figure S7.The effect of mean date (expressed as the number of days since 1st April) on the relative abundance index of song thrush (difference in abundance index between 2018 and 2017).Estimates from the generalized additive mixed model with Gaussian error distribution.Total survey duration (in minutes) was included as the offset variable.Ribbons are 95 % confidence intervals.Explanations: ** -p < 0.01.

Figure S8 .
Figure S8.Scheme of the used broadcasts of the attractive (a), repulsive (b), and mixed (c) social information.Abbreviations used: STsong thrush, Ggoshawk, A, B, C, Dsound samples of particular individuals.

Table S1 .
The mean characteristics (area and isolation index) of the studied forest patches across the experimental groups.Forest area is in hectares and forest isolation index is the nearest-neighbour distance in meters.Explanations: SD -standard deviation, Min -minimal value, Max -maximal value.

Table S2 .
Test of differences in patch area among forest patches assigned to different treatments.Results of the general linear additive model with broadcast type as a factor and interaction between geographic coordinates as a smooth term.Explanation: CI -95% confidence interval.