Falconiformes

Pelegrin, Jonathan S, Cantalapiedra, Juan L, Gamboa, Sara, Menéndez, Iris & Hernández Fernández, Manuel, 2023, Phylogenetic biome conservatism as a key concept for an integrative understanding of evolutionary history: Galliformes and Falconiformes as study cases, Zoological Journal of the Linnean Society 198 (1), pp. 47-71 : 49-52

Falconiformes

The order Falconiformes (falcons and allies) includes 66 species classified within the family Falconidae ( Fuchs et al., 2012, 2015; Del Hoyo & Collar, 2014). Overall, falcons are faunivorous, with a body size spectrum that varies from small species, such as the black-thighed falconet, Microhierax fringillarius Drapiez, 1824 , near 35 g, to the gyrfalcon, Falco rusticolus Linnaeus, 1758 , weighing ~ 1.7 kg ( Ferguson-Lees & Christie, 2001). Ecologically, falcons dwell in a great variety of habitats, including forest, savanna, desert and steppes, with many species showing a wide spectrum of environmental occupation, as is the case for the peregrine falcon, Falco peregrinus Tunstall, 1771 , with a worldwide distribution ( Wilcox et al., 2019). The evolutionary relationships of falcons are well established on molecular evidence ( Jarvis et al., 2014; Fuchs et al., 2015; Prum et al., 2015; Kimball et al., 2019; Wilcox et al., 2019). Only one family is recognized in the order ( Falconidae ), and two subfamilies are well established: Herpetotherinae (laughing and forest falcons) and Falconinae (caracaras, falconets, falcons and kestrels). For our purposes, we used the complete falcon phylogeny published by Fuchs et al. (2015), which includes all the species. This phylogeny was calibrated with the Early Miocene Thegornis F. Ameghino, 1895 from Argentina, which is considered a sister taxon of Herpetotherinae ( Noriega et al., 2011), and the Middle Miocene Pediohierax Becker, 1987 , an early North American member of Falconinae ( Becker, 1987) . Finally, we followed Del Hoyo & Collar (2014) for taxonomic consistency in all scientific names and the nomenclature of avian clades.

BIOME DATA

We selected the biome classification system developed by Walter (1970) and modified by Hernández Fernández (2001), which establishes ten biomes, considering the annual relative distribution of temperature and precipitation ( Table 1). We reviewed diverse palaeoclimatic studies that reveal variation in the age and spatial dynamics of the biomes ( Pennington et al., 2004a; Jetz & Fine, 2012; Mucina, 2019; Landis et al., 2021). The oldest biomes, such as evergreen tropical rainforest, have been reported since the Cretaceous and the Palaeocene ( Davis et al., 2005; Jaramillo et al., 2010; Jaramillo & Cárdenas, 2013; Eiserhardt et al., 2017; Carvalho et al., 2021). During the Eocene, owing to climatic fluctuations, areas of tropical deciduous woodland appeared ( Bredenkamp et al., 2002; Jacobs, 2004; Pennington et al., 2004a; Jaramillo & Cárdenas, 2013), which extended widely during the Oligocene and Late Miocene ( Pennington et al., 2004b; Edwards et al., 2010; Charles-Dominique et al., 2016). Likewise, during the Eocene, temperate evergreen forest and broadleaf deciduous forest biomes in high-latitude zones emerged ( Axelrod, 1966; Markgraf et al., 1995; DeVore & Pigg, 2013; Landis et al., 2021). After the Mid-Miocene Climatic Optimum, the reduction in global temperature triggered the increase of arid conditions in different areas ( Zachos et al., 2001; Charles-Dominique et al., 2016; Harzhauser et al., 2016; Hurka et al., 2019). These factors promoted the origin of boreal coniferous forests in the highest latitudes ( Pound et al., 2011; Popova et al., 2012, 2017). Meanwhile, open biomes, such as savannas, steppes and deserts, began to expand in the arid and semi-arid regions ( Axelrod, 1985; Bredenkamp et al., 2002; van Dam, 2006; Byrne

Abbreviations: F, forest environments; O, open environments.

et al., 2008; Senut et al., 2009; Guerrero et al., 2013; Charles-Dominique et al., 2016; Hurka et al., 2019). Pliocene and Pleistocene geological and climatic events caused phases of icehouse and subsequent glaciations, which made possible the emergence of biomes such as the sclerophyllous woodland–shrubland in subtropical latitudes ( Hernández Fernández et al., 2007; Buerki et al., 2012) and the tundra near the poles ( Zimov et al., 1995; Hewitt, 2003). Moreover, this period implied the prominent expansion of steppes in high latitudes ( Franzke et al., 2004; Kahlke, 2014) and deserts in the subtropical belts ( Bobe, 2006; Senut et al., 2009), with the opposite effect of contractions in the area of tropical and wet biomes ( Pennington et al., 2000; Hooghiemstra & Van der Hammen, 2004; Jaramillo & Cárdenas, 2013; Raes et al., 2014; Dexter et al., 2018; Table 1).

The biome occupation of each species was determined following the methodology developed by Hernández Fernández (2001). According to the geographical distribution of each species, we codified the presence or absence of the species in each biome, considering the overlapping and relative size of its reported geographical range ( Del Hoyo & Collar, 2014; Birdlife International, 2019). If ≥ 15% of the geographical range of the species was situated within a specific biome, the species was considered to occupy that biome. For the case of biomes with relatively small areas in relationship to the geographical range of the species, we also recorded the presence of species when the species inhabited ≥ 50% of one climatic dominion, that is, the south-eastern coastal forests of South Africa for Numida meleagris Linnaeus, 1758 . Species with presence in mountain environments were also recorded as present in the analogous biome, considering the elevational climatic gradient ( Hernández Fernández & Vrba, 2005).

ANCESTRAL BIOME RECONSTRUCTION

In order to trace ancestral biome occupation, we modelled the occurrence of species in the ten biomes along the evolutionary history of Galliformes and Falconiformes , taking a phylogenetic approach. The ancestral biome occupation for all lineages was estimated through maximum likelihood analysis of geographical range evolution using the package BIOGEOBEARS (BioGeography with Bayesian and Likelihood Evolutionary Analysis) implemented in R ( Matzke, 2015). This analysis allowed us to model the dynamics of biome occupation in relationship to the timing of cladogenesis based on the splitting times of the phylogenetic trees. Although it was designed originally for the study of biogeographical evolution, BIOGEOBEARS is an analytical approximation that allows probabilistic inference of ecological characters, biome occupation in our case, integrating different models onto a time-calibrated phylogenetic tree ( Batalha-Filho et al., 2014; Lynch Alfaro et al., 2015; Buckner et al., 2015; de Medeiros & Lohmann, 2015). This tool enabled us to infer the ancestral biome in each node along the phylogeny by considering biomes as geographical areas with their own connectivity dynamics according to the climatic history of the Earth ( Cardillo et al., 2017; Landis et al., 2021). BIOGEOBEARS estimates maximum likelihood for ancestral states during speciation events, modelling the transitions between different states (biomes occupied) along the phylogenetic branches as a function of time. The analyses were conducted using the dispersal–extinction–cladogenesis (DEC) model ( Ree & Smith, 2008), modified in BIOGEOBEARS with the +J parameter (for jump), which models the process of founder-event speciation ( Matzke, 2014a, b, 2015). We performed the DEC model a priori because it includes all biogeographical processes available in the DIVALIKE and BAYAREALIKE models. We conducted a BIOGEOBEARS model considering founder events (parameter +J) because of the long-distance dispersal capacities expected in birds ( Price & Clague, 2002; Pigot & Tobias, 2015; Hosner et al., 2017), which are expected to be important in the context of an island system, such as the one studied here, based on multiple climatic dominions. This is especially so, considering the relevance of transcontinental colonization events for avian diversification. Specifically, the DEC and DEC+J models have been consistent to explain colonization processes in biogeographical and macroevolutionary studies carried out in various lineages of birds, including Megapodidae ( Harris et al., 2014) , Motacillidae ( van Els et al., 2019) , Coraciformes ( McCullough et al., 2019), Trogoniformes ( Oliveros et al., 2020), Rallidae (García-R & Matzke, 2021) and Sphenisciformes ( Pelegrin & Acosta Hospitaleche, 2022). According to these studies, we consider that the +J parameter could be relevant in the evolution of Falconiformes and Galliformes because many lineages have achieved remarkable dispersion (e.g. Falco eleonorae Gene, 1839 has colonized Madagascar; several Megapodius Gaimard, 1823 species are distributed across islands of Micronesia), sometimes associated with long-distance migration (e.g. Falco naumanni Fleischer, 1818 or Coturnix coturnix Linnaeus, 1758 ). Finally, although Ree & Sanmartín (2018) criticized the conceptual basis of the DEC/ DEC+J model, Klaus & Matzke (2020) rejected their claims based on standard, widely accepted principles of model evaluation and comparison (for details, see Klaus & Matzke, 2020).

Processes of climate change have altered the ecological conditions of the Earth over geological time. This has led to different biomes having particular histories in relationship to their origin and processes of geographical fluctuation. Therefore, in our reconstruction we incorporated information about the availability of each biome along the Cenozoic by implementing ‘biome existence–connection matrices’ (Supporting Information, Table S1). These matrices were established according to anextensivereviewofpalaeoecologicalandpalaeoclimatic literature (for references, see Table 1) and allowed us to reflect the presence or absence of biomes since the Eocene/Oligocene (time of origin of the study groups, according to the phylogenies) until the Pleistocene– Holocene. These matrices also reflect the connectivity among biomes, hence their potential colonization pathways. For example, there is extensive documentation on the climatic and geographical relationship between the evergreen tropical rainforest and tropical deciduous woodland biomes ( Dexter et al., 2018) that supports the plausibility of colonization processes between them, whereas the rainforest and tundra have never had direct connections. This methodological approach allows us to introduce temporal constraints related to the geological, climatic and ecological history of biomes in the analysis. Likewise, there is some uncertainty regarding the timing of the onset of the biomes ( Jetz & Fine, 2012). For this reason, we constructed two biome existence–connection matrices based on geological and fossil evidence in a broad time interval (see references in Table 1). These biome matrices were established as follows, one for the Oligocene–Miocene (in which biomes III, IV, VII and IX were not yet present) and a second one for the Pliocene– Pleistocene (with all extant ten biomes). The connectivity between two biomes was coded with ‘1’ and the absence of connectivity among biomes or non-existence was coded as ‘0’ (Supporting Information, Table S1).

To avoid the computational intractability of the analysis with all 1023 potential biome combinations ( Hernández Fernández&Vrba, 2005) in two phylogenies with a high number of species, we complemented the biome existence–connection matrices with information on biome occupation in recent birds. To some extent, extant avifauna shows potential biome occupation in birds, considering biological clade properties such as flight capability and migratory behaviour, which have an influence on dispersal ability ( Rolland et al., 2014). Thus, we reduced these 1023 potential biome combinations to a set of only 239 biome combinations, which represented the observed combined biome occupations in all extant non-passerine birds (3951 spp.; Supporting Information, Table S2).

The process to reconstruct the ancestral biome occupation of Galliformes and Falconiformes combined the information on living species biome occupation, tree topology, availability and connections among biomes through time, at the same time yielding the likelihood values for each biome combination in each node. The BIOGEOBEARS output for a ten-area analysis was complex. In order to organize the results and obtain general patterns of historical biome occupation, we established a criterion based on selection of the biome combinations with the highest probabilities for each node, which were organized in a decreasing fashion according to their likelihood value.The probability scores for these biome combinations were cumulative until a value of 0.5 was reached. The relative likelihood scores of each biome within the selected combinations were summed, and all biomes with a cumulative likelihood of> 0.25 were retained for each node (Supporting Information, Tables S3 and S4). For example, if, for a given node, the most probable biome combinations were III–IV (presence in desert and sclerophyllous woodland), with a likelihood value of 0.4, and IV–VII (presence in sclerophyllous woodland and steppe), with a likelihood value of 0.12, this would imply that the values of relative probability for each biome were defined as follows: III = 0.4, IV = 0.52 and VII = 0.12. We would then consider that only biomes III and IV were reconstructed robustly for that particular node.

NICHE CONSERVATISM AND BIOME COLONIZATION

To assess the phylogenetic biome conservatism patterns for both Galliformes and Falconiformes , we counted the transitions between biomes from each node to its descendant nodes. In this sense, the transition was defined, for each biome, as the path between an ancient node with respect to its descendant node. The transitions could be conservative (i.e. the biome state was preserved between ancestor and descendent), contributing to phylogenetic biome conservatism, or non-conservative, implying the colonization of a biome different from those occupied by its ancestor ( Fig. 1). Also, we counted the cases of biome loss between the biome occupation state of the ancestor and descendent. The reconstruction of biome occupation for each node of the phylogeny allowed us to describe the occupation trends of the ten different biomes through time. We counted the transitions, comparing the state in an ancestral node with respect to its descendants, and codified whether a specific biome was conserved or not (implying change or loss) between ancestor and descendent nodes. This approach to conservatism quantification was similar to the one proposed by Duchêne & Cardillo (2015) based on latitudinal zones.

Numerous works have focused on the related hypothesis of tropical conservatism, which aims to explain the observed imbalance in diversity richness in tropical latitudes compared with temperate zones. This hypothesis proposes that the great species richness found in tropical environments is attributable to the niche conservatism of the lineages linked in their origin to these environments, which expanded across the Palaeogene ( Wiens & Donoghue, 2004). However, the tropical conservatism hypothesis has