Defining biogeographic regions
To define a practical biogeographical categorization of regions, we balanced capturing the complexity of our focal regions (i.e. the islands between Australia and continental south-east Asia), with limiting biogeographic model complexity (i.e. minimising states for which there is insufficient data to estimate parameters). We also structured regions to a) allow comparisons between the major continents and islands with long histories of isolation from continental landmasses, and b) to allow some investigation of biogeographic patterns within the islands of the geologically complex IAA.
To begin with, we used 21 sub-regions or areas to describe insular and continental region pigeon diversity (Table S3). Geographic range data was initially summarised from the Handbook of the Birds of the World (Baptista et al. , 1997) and then cross-referenced and updated based on two online databases (https://www.iucnredlist.org/ and https://www.worldbirdnames.org/new/). Grouping of areas was initially explored using turnover of phylogenetic diversity (phylobeta: Graham & Fine 2008; Leprieur et al. 2012). Phylogenetic beta-diversity was estimated using Simpson’s phylobeta index (pβsim) using the BEAST MCC tree (after removing outgroups), and compiled into a pairwise pβsim distance matrix (using custom script GTREER5) (see Dryad repository). Using this distance matrix we visualised differences between regions using both hierarchical clustering (average = UPGMA) and Multi Dimensional Scaling (MDS) 2-D plot (hclust and cmdscale, ’stats’ R package; R Core Team 2016)). Five taxa that occur in >8 sub-regions were removed from estimation of pβsim to limit emphasis on recent connections at the expense of underlying older endemicity, which is the focus of this study.
We then assigned taxa to broad regional categories (hereafter referred to as regions) based on modifying the pre-existing historical regionalizations (e.g. Jønsson et al. 2010, and also including the IOC9.2 breeding range zones) with information from both phylogenetic turnover (Fig. S1) and geological models for the history and formation of islands around the Indo-west Pacific (see above and Zahirovic et al. 2016). This resulted in an eight region scheme: 1) New World – North, South and Middle America: 2) – Old World (Europe, Africa and mainland Asia combined); 3) – Philippines; 4) Wallacea – Sulawesi and the Lesser Sundas; 5) West Melanesia - New Guinea plus the islands of Maluku; 6) Indian Ocean Islands – especially the Mascarenes and Madagascar; 7) Australia (not including New Guinea) and; 8) Pacific Ocean – the islands of the south-west Pacific including East Melanesia, New Caledonia and New Zealand. Assignment of phylogenetically sampled species to these eight regions is listed in Table S4. We note that our configuration of Wallacea does not include Maluku. This is because our phylogenetic turnover analyses indicate that much of Maluku is more closely allied to New Guinea (Fig. S1A). Geological (Hill & Hall 2003) and biological (Oliver et al. , 2022) data also indicate a close relationship between islands of Maluku and New Guinea. We separated Australia from West Melanesia in our analyses because: a) phylobeta analyses indicate the dominant pigeon faunas of the two regions are relatively discrete and b) geological models indicate that the New Guinea has been separated from the main Australian landmass by at least a shallow sea for most the late Cenozoic (see geological context above). Species that occur across regions were scored as such and allowed to have multiple states in analyses.
To investigate the role of islands in pigeon diversification and dispersal, we collated regions into broader categories of islands (a composite of the Wallacea, West Melanesia, Philippines and Pacific Ocean regions) versus continents (Americas, Old World and Australia). This was done by post hoc combining states and state changes inferred in the full eight region analyses into summary results (see below).