Computational protocol: Family living sets the stage for cooperative breeding and ecological resilience in birds

Similar protocols

Protocol publication

[…] We collected data on the social system, life history, and ecological parameters of bird species from the literature (see ). We used 3 different criteria to differentiate between the different social systems, using the known duration of family associations (i.e., the time offspring remain with parents beyond nutritional independence []), the occurrence of family groups during the non-breeding season (when the exact time offspring remain with parents beyond independence was unknown), or the occurrence of cooperative breeding [] and the kin relationship of helpers [] (see ). We did not categorize occasional cooperative breeding species as cooperative breeders [] (based on the first 2 criteria above). Occasional cooperative breeding resembles interspecific feeding, in which individuals feed offspring of another species, and thus, different factors are likely to be associated with occasional cooperative breeding and regular cooperative breeding [].Species were categorized as sedentary (maximally engage in local movements) or migratory (short-, long-distance, and altitudinal migrants). Species that only use 1 food category were categorized as food specialists, whereas species that used at least 2 different food types were categorized as food generalists (see for details on the food categorization). Habitat openness was calculated based on aerial images, following the International Union for Conservation of Nature (IUCN) Habitat Classification Scheme []. Nest type was categorized as a binary variable (cavity breeders: nests in cavities, cliffs, and caves; other nests: all other nest constructions). We used the mean body weight (combining male and female weight) and distinguished precocial from altricial species (categorizing semiprecocial species as precocial and semialtricial species as altricial).Climatic variables were computed from data provided by the Climatic Research Unit Time Series 3.21 database at the University of East Anglia (http://catalogue.ceda.ac.uk/uuid/ac4ecbd554d0dd52a9b575d9666dc42d; downloaded 7 April 2014) and NASA (http://neo.sci.gsfc.nasa.gov/view.php?datasetId=MOD17A2_M_PSN; downloaded 5 December 2013). We calculated for each species:Precipitation: mean, within and between year variance, predictability during the whole year;Precipitation: mean, within and between year variance during the MGS only;Temperature: mean, within and between year variance, predictability during the whole year;Temperature: mean, within and between year variance during the MGS only;NPP: mean, variance, predictability during the whole year;NPP: mean, within and between year variance, predictability during the MGS only;MGS length;Habitat heterogeneity (according to []).Since the variance often increases with the mean (i.e., Taylor’s law []), it has been suggested that the coefficient of variance may be a more appropriate measurement to assess climatic variability. Thus, we re-ran our analyses using the coefficient of variance where appropriate (for variables measured on absolute scales, i.e., precipitation, temperature). Both the PCA () and the multinomial model () resulted in qualitatively similar results as our main analyses, indicating that our choice of variability metric did not bias our results.All statistical analyses were performed in R with the packages Diversitree [], phytools [], and MCMCglmm []. Ancestral state estimation was performed using the MuSSE function of the Diversitree package [] on a consensus phylogeny estimated from a sample of 1,000 phylogenetic trees [] with the maximum parsimony matrix method using the Hackett tree backbone. We note that using a consensus phylogeny with the Ericsson backbone returned qualitatively identical results. Also, using a model in which speciation and extinction rates were allowed to vary resulted qualitatively in the same results as the main models with diversification rates fixed to be equal across breeding modes (). We fitted phylogenetically controlled multinomial models to our data using MCMCglmm []. The response variable in all models was a categorical representation of social system (3 nominal levels: non-family living, family living, and cooperative breeding). The phylogenetic random effect was modelled based on a recent phyla-wide phylogeny []. To account for the uncertainty of phylogeny estimation, we refitted the main model with 50 randomly selected trees from the posterior distribution of trees published in Jetz et al. []. Given that ancestral character reconstruction may be biased when characters influence diversification [], we also used a phylogenetic controlled PCA (phyloPCA function in phytools) [], resulting is a somewhat different PC structure (). However, running our main model with this PC resulted qualitatively in the same results (). […]

Pipeline specifications

Software tools Diversitree, Phytools
Application Phylogenetics
Organisms Danio rerio