Biogeography_analyses

To investigate the factors that may have influenced the distribution of <i>Ocymyrmex</i> species in southern Africa, we inferred their biogeographic histories using the R package BioGeoBEARS (BioGeography with Bayesian (and likelihood) Evolutionary Analysis in R Scripts) (Matzke, 2018). This likelihood-based model is designed to perform inferences of biogeographical histories on phylogenies and to compare different models of range evolution (Matzke, 2013a). The following biogeographical models were tested against our data to estimate the ancestral area of <i>Ocymyrmex</i> species: a Dispersal-Extinction-Cladogenesis (DEC) model, which allows the geographic range to change across a phylogeny through several events (Ree, 2005; Ree and Smith, 2008); a likelihood version of the Dispersal-Vicariance model (DIVALIKE) (Ronquist, 1997); and a likelihood version of the Bayesian Analysis of Biogeography (BAYAREA-LIKE), which is a Bayesian model that samples geographical history along phylogenetic branches, together with sampling parameter values (Landis et al., 2013). These models have free parameters specifying the rate of dispersal (<i>d</i> = range expansion) and extinction (<i>e</i> = range of contraction)<i> </i>along the phylogeny branches. At cladogenesis events, DEC assumes that one daughter lineage will always inherit the ancestral range if the ancestor lives in a single area; if the ancestor is widespread, one daughter lineage will live in a subset of that area, or one area will split off by vicariance. By contrast, DIVA only allows vicariance. BayArea assumes that no range evolution occurs at cladogenesis, and therefore the ancestral range is copied by both daughters (Matzke, 2013b). A founder event speciation model (+j) was also tested in combination with each base model.BioGeoBears requires a time calibrated tree and geographic distribution units as input files. <i>Ocymyrmex </i>distribution records were extracted from Iziko Museum’s Specify6 database (Specify version 6.7.01; Specify Collections Consortium | Software for Biological Collections and Samples). Given the tight habitat affinity of <i>Ocymyrmex</i>, species occurrences were assigned and coded according to the biomes defined by Dinerstein et al. (2017) (Appendix B: Table B2). The following biome regions were designated and used as biogeographical units: Tropical &amp; subtropical moist broadleaf forests; Tropical &amp; subtropical grasslands, savannas &amp; shrublands; Flooded grasslands &amp; savannas; Montane grasslands &amp; shrublands; Mediterranean forests, woodlands &amp; scrub; and Deserts &amp; xeric shrublands. The dated phylogeny produced by MCMCTREE was pruned to exclude outgroups and was used as an input file. The Akaike information criterion (AIC) was used to select the best-fitting model. The maximum number of areas any species may occupy was set to five, as the <i>Ocymyrmex </i>species with the largest distribution range, <i>O. flaviventris</i>, can be found in five biomes. We followed guidelines in http://phylo.wikidot.com/biogeobears.