Eighty-three lineages that took over the world: a first review of terrestrial cosmopolitan tetrapods

Authors


Correspondence: Şerban Procheş, Discipline of Geography, University of KwaZulu-Natal, Westville Campus, Private Bag X54001, Durban, 4000, South Africa.

E-mail: setapion@gmail.com

Abstract

Aim

While there is a huge macroecological and biogeographical literature addressing endemism, very little has been done to systematically study lineages that are widely distributed across the globe. Our aim here was to list and analyse those lineages of terrestrial tetrapod vertebrates found in 60–90% of the world, loosely termed here as cosmopolitan.

Location

Global.

Methods

Two sets of geographical units and three occupancy criteria were used to list, analyse and map cosmopolitan lineages and their sister lineages.

Results

Among the 83 lineages identified, 2 were represented by amphibians, 9 by reptiles, 13 by mammals, and the remainder by birds, of which 12 were passerines and 47 were non-passerines. All these lineages are present in parts of Southeast Asia, most of them throughout much of Eurasia and Africa, but fewer in South America and very few in Australia. Only three of the lineages (all reptiles) are likely to exemplify vicariance or early dispersal-driven cosmopolitanism, the rest having attained world-wide distribution via extensive, geologically recent dispersal. The distribution of sister lineages indicates that many cosmopolitan lineages probably originated in the savanna regions of Africa, some in Southeast Asia, and fewer in tropical America.

Main conclusions

Cosmopolitan distributions in tetrapods are primarily the result of dispersal, with large body size and the ability to fly being two key correlates of rapid global colonization. We argue that a cosmopolitan lineage framework in biogeographical and ecological studies could add great depth to the understanding of evolutionary success, and would be highly relevant to the field of invasion biology.

Introduction

The field of macroecology relies heavily on the concept of range size, which measures how widely distributed a taxon or lineage is (Gaston & Blackburn, 2000). The efforts of ecologists focus largely on narrowly distributed taxa, that is, on only one end of the range-size spectrum (Gaston, 1994). While this focus is very much justified from a conservation perspective, it is less easy to justify the lack of a substantial contemporary body of literature pertaining to the opposite end, namely naturally very widespread taxa or lineages. Indeed, although these taxa are typically not of conservation concern, an understanding of their success would be relevant to the growing field of invasion biology (Stohlgren et al., 2011), and furthermore of great interest in telling the story of life on Earth, in particular where range expansion meant colonizing the world's less hospitable environments and isolated landmasses.

The phrase ‘cosmopolitan taxa’ describes an old intuitive concept, referring to organisms spread globally, or nearly so. De Candolle (1855) was possibly the first to use the term cosmopolitan in a biogeographical sense. Only recently have there been attempts to define this phrase in an evolutionary perspective (for example Hoffmann, 1996, considers cosmopolitan cyanobacteria to include those groups in which dispersal happens more rapidly than evolution) or to model cosmopolitan distributions (Goldberg, 2007, looking at the age of endemic versus cosmopolitan lineages). Lists of cosmopolitan (or ‘widely distributed’) organisms usually refer to taxa such as species, genera or families (e.g. Darlington, 1957; Good, 1964; Fenchel & Finlay, 2004). However, in a broad evolutionary perspective, one should of course be looking at lineages. Every organism belongs in fact to a cosmopolitan lineage, the question being how inclusive that lineage needs to be to meet a given criterion for being classified as cosmopolitan. For example, the osprey (Pandion haliaetus) can be described as a cosmopolitan species; on the other hand, groups such as shrews (Soricidae) are also cosmopolitan, but it may take the entire family to give them this status, and the family includes hundreds of species, most of which have narrow geographical ranges. Accordingly, the tree of life can be subdivided into cosmopolitan groups (groups defined so as to be minimally inclusive while still meeting a given criterion for being called cosmopolitan) and grades of less widely distributed lineages leading up to these.

Recently, a number of case studies have investigated the historical biogeography of specific cosmopolitan lineages using molecular methods (Richardson et al., 2004; Funk et al., 2005; Bossuyt et al., 2006; Lovette & Rubenstein, 2007; Pramuk et al., 2008; Voelker et al., 2009). These studies often indicate the successive colonization of multiple regions, and in some cases remarkable long-distance dispersal events. This raises the following question. Are all or most cosmopolitan lineages the result of dispersal from a centre of origin, or can at least some of them be explained by vicariance, with lineages surviving in multiple landmasses following the break-up of Pangea?

If dispersal is the only (or main) explanation, then organisms with better dispersal abilities should be more likely to attain cosmopolitan status rapidly, and consequently these groups should require fewer species to make a lineage cosmopolitan. Furthermore, for lineages that are found in most parts of the world but not in the remaining few, these areas of absence could be relevant. Under the dispersal scenario, the proportion of lineages that are cosmopolitan in more isolated landmasses is likely to be different from that in well-connected ones. Absence from areas with limiting environmental factors (aridity, cold climates), on the other hand, would not differentiate between vicariance and dispersal scenarios.

Under the dispersal scenario, the centre of origin for each cosmopolitan lineage can be sought by examining the distribution of the lineage that is sister to the cosmopolitan one, and if many such sister lineages occur in the same restricted regions, this could be indicative of a repetitive process of global colonization starting from these regions. This pattern has often been implied in phrasings such as ‘out of Africa’, ‘out of Australia’ or ‘out of Asia’ (Waters & Roy, 2004; Braby & Pierce, 2007; Donoghue, 2008), although no cross-lineage tests for such hypotheses have focused specifically on cosmopolitan lineages.

In this study, we use geographical and phylogenetic mapping of cosmopolitan lineages in tetrapod vertebrates (chosen as a study group by virtue of the availability of good distributional data) to answer these questions. First, in a descriptive effort, we list all cosmopolitan terrestrial tetrapod minimum-spanning lineages. Second, we perform phylogenetic and geographical mapping exercises, testing the following predictions: (1) regions that were historically isolated for long periods of geological time, such as Australia and South America, have a low representation of cosmopolitan clades (and thus more ‘original’ vertebrate faunas); (2) regions with extreme environmental conditions (either low temperatures or low precipitation), as well as recent and isolated oceanic islands, contain a greater proportion of cosmopolitan lineages (less ‘original’ faunas); (3) cosmopolitan lineages originate more often in certain regions of the world, irrespective of their intrinsic characteristics (the ‘out of…’ hypothesis); and (4) the number of species in cosmopolitan lineages is influenced by average body size and the ability to fly. Third, we discuss these results in the light of the available lineage age data.

Materials and Methods

Three criteria based on two data sets were employed in deciding what lineages can be considered cosmopolitan. Initially, in an exploratory phase, relevant distributional data sources (Nowak & Paradiso, 1983; Halliday & Adler, 1986; Del Hoyo et al., 1992 onwards; Perrins, 2009; Roberson, 2009; AmphibiaWeb, 2010; Frost, 2010; Uetz & Hallermann, 2010) and phylogenetic references (see Appendix S1 in Supporting Information) were browsed, and the distribution of monophyletic lineages was mapped for equal-area cells (identical in coverage to those used in the worldmap software; Williams, 1999). After preliminary trials, mapping was restricted to those cells located between 60° N and 60° S that contain at least 10% land. This resulted in 281 cells (Appendix S2). In order to approximately match the occupancy levels used in previous, intuitive approaches, lineages were considered cosmopolitan if present in at least 215 of these 281 cells (76.5%) (criterion a). If a lineage fell short of meeting this criterion, further sister lineages were included, until either the criterion was met, or a node in the tree was reached where the sister lineage was already a cosmopolitan lineage. The lineages thus identified as cosmopolitan or nearly so were then mapped using the WildFinder data set (World Wide Fund for Nature, 2010), a comprehensive presence/absence data set for tetrapod vertebrate species in 821 terrestrial biogeographical subdivisons of the world, termed ‘ecoregions’, intuitively delimited by Olson et al. (2001). Based on the distributional information in this data set, lineages were considered cosmopolitan if present in at least 500 (60.9%, criterion b) or at least 600 (73.1%, criterion c) of the 821 ecoregions. While it can be argued that the geographical units in each of the two data sets are less than perfect (unequal dry land area in both cases, incomplete distributional data in mapping lineages in the former, and intuitive predefined boundaries for the latter), the use of both gave us some assurance that our results were not simply driven by drawbacks in the way the distributional data were assembled. The WildFinder data conveniently allowed us to calculate lineage distribution and diversity, and to map lineages in ArcGIS (ESRI, 2008) within a manageable number of units. Consequently, lineages meeting any of the three criteria are listed and used in the phylogenetic mapping, but only lineages identified using criteria (b) and (c) [thus using World Wild Fund for Nature's (WWF) WildFinder data] are used in the geographical mapping exercises.

To assess the representation of cosmopolitan lineages in the WWF ecoregions, we first visually inspected the distributional maps produced for individual lineages, looking for patterns indicative of historical and environmental limitations. We then calculated the number of lineages present in each ecoregion; next the numbers of species in all cosmopolitan lineages were added for each ecoregion; and finally this sum was divided by the total number of tetrapod species in that ecoregion to give a measure of proportional representation.

To assess the geographical origin of cosmopolitan lineages, we assumed that this would most often be the area where closely related but less widespread lineages occur (cf. for example Erwin, 1985). In order to identify this area, one should ideally consider the distribution of multiple lineages diverging from the cosmopolitan lineage as its occupancy increases. Owing to the lack of a fully resolved phylogenetic tree, here the sister lineage of each cosmopolitan lineage was identified and mapped. To identify world regions where multiple cosmopolitan lineages originated, we calculated (1) the total number of sister lineages, (2) the number of range-restricted lineages (sister lineages present in fewer than 120 ecoregions, this value being based on the most notable inflection point in a histogram of sister lineage range sizes; graph not presented here), and (3) the total number of species belonging to sister lineages present in each ecoregion.

The total number of sister lineages was calculated (i) unweighted, as well as weighted by range size, which was achieved by dividing each presence by the total number of ecoregions where that specific sister lineage was present; the number of ecoregions being (ii) untransformed, (iii) square-root-transformed, or (iv) log-transformed (see Williamson & Gaston, 1999). This weighting was aimed at giving greater value to those ecoregions in which multiple range-restricted sister lineages occur.

In assembling information on phylogenetic relationships, sister lineages were determined from the most detailed source in each case. Where two distinct phylogenetic studies were available, the more recent one was preferred, and where the studies were less than 3 years apart, we chose the tree with higher support values. Where more than two phylogenetic studies were available, we followed the one presenting results closest to a putative cross-study consensus tree. In the very few cases where no phylogeny was available, we used taxonomy as a surrogate (see Appendix S1 for reference details). We considered only those references published (at least online) by October 2011, when analyses and mapping were performed.

To assess the importance of body mass and flight in attaining cosmopolitan status, the numbers of species needed to make a lineage cosmopolitan (according to criterion b) were compared across the following categories: (1) flight versus no flight (all bird and bat lineages listed here are dominated by species that have the ability to fly), and (2) < 10 g; 10–100 g; 100 g–1 kg; > 1 kg, using spss 21.0 (IBM Corporation, 2012). Body mass for a lineage was recorded as the estimated cross-species average for adult body mass for all species in the lineage, derived from the same references as the geographical distribution [listed under criterion (a) above].

Results

Eighty-three lineages were identified that met at least one of the three criteria for being cosmopolitan. Of these, only 24 belonged to groups other than birds (Fig. 1, Table 1). Cosmopolitan lineages were particularly numerous in water birds (waders, herons) and raptors; in some of these cases, such lineages were represented by single species (e.g. great egret, osprey, peregrine). Of these 83 lineages, 77 were cosmopolitan according to criterion (b), occurring in at least 500 ecoregions. When increasing this number to 600 (criterion c), 17 lineages remained cosmopolitan in their original form, 28 had to be expanded by incorporating further sister lineages (in four cases, these included other cosmopolitan lineages), and the rest were disqualified (see Table 1 for cosmopolitan lineages; Appendix S1 for sister lineages).

Table 1. Listing of the cosmopolitan terrestrial tetrapod vertebrate lineages mapped on the phylogeny in Fig. 1, indicating whether they fulfil criteria (a), (b) and/or (c) for being considered cosmopolitan
LineageCriterion (b)Criterion (c)
NumberNameDescriptionNumber of ecoregions occupiedNumber of speciesDescriptionNumber of ecoregions occupiedNumber of species
  1. In the cases of criteria (b) and (c), the identity, occupancy and number of species for each lineage are indicated. Lineages not cosmopolitan under criterion (b) are cosmopolitan only under criterion (a). Asterisks refer to criterion (a), as follows: * lineage cosmopolitan only if more inclusive than under (b); ** lineage cosmopolitan even if less inclusive than under (b); *** included in the lineage above; **** not a cosmopolitan lineage.

1Common frogRanoidea701951As for (b)701951
2Toad*Core Bufonidae561201Bufonidae, Dendrobatidae, Leptodactylidae, Centrolenidae, Allophrynidae6031478
3ShrewSoricidae562312Not cosmopolitan  
4Free-tailed bat****Molossidae54898Not cosmopolitan  
5Mouse-eared batCore Myotis59997Myotinae, Kerivoulinae, Murininae637134
6Serotine batNycticeiini51044Nycticeiini, Pipistrellini714181
7Barbastelle*Lasiurini, Plecotini52331Not cosmopolitan  
8HareCore Leporidae55854Not cosmopolitan  
9MouseMuridae, Cricetidae7151130As for (b)7151130
10SquirrelCore Sciuridae561271Sciuridae, Aplodontiidae, Myoxidae601303
11DeerCervidae, Moschidae, Bovidae639187As for (b)639187
12Cat*Core Felinae52316Core Felidae60630
13WolfCanini56617Canidae60233
14Weasel****Mustelini51023Core Mustelidae60246
15OtterLutrini50912Included in above  
16Gecko*Geckonidae516695Not cosmopolitan  
17SkinkScincidae5881224Not cosmopolitan  
18Common lizardLaterata501962Not cosmopolitan  
19Chameleon Iguania 5551021Not cosmopolitan  
20AdderViperidae512237Not cosmopolitan  
21Cobra****Elapidae, Lamprophiidae549844Not cosmopolitan  
22Grass snake****Natricidae521516Colubridae s. str., Natricidae6151188
23Smooth snakeCore Colubridae s. str.572577Included in above  
24TurtleTestudines528286Not cosmopolitan  
25Partridge*Core Phasianidae566159Not cosmopolitan  
26GooseNot cosmopolitan  Not cosmopolitan  
27ShovellerAnas clypeata group6089As for (b)6089
28Mallard***Anas platyrrhynchos group5128Core Anas62120
29DoveColumbinae60777As for (b)60777
30Dabchick* Tachybaptus, Podilymbus 6176As for (b)6176
31SwiftApodini56427Apodidae70491
32Nightjar*Chordeiles, Caprimulgus and related genera62241As for (b)62241
33CuckooCuculiformes701141As for (b)701141
34Crake*Core Porzana, Crex50010Not cosmopolitan  
35CootNot cosmopolitan  Core Gallinula, Fulica63917
36MoorhenCore Gallinula5254Included in above  
37Cormorant**Phalacrocoracidae, Anhingidae, Sulidae51844Not cosmopolitan  
38IbisNot cosmopolitan  Not cosmopolitan  
39Night heron**** Nycticorax nycticorax 5031Not cosmopolitan  
40Bittern*Core Ixobrychus5267Core Botaurinae64212
41Little egret****Core Egretta5493 Egretta, Ardeola 60917
42Great egret Chasmerodius 5631Not cosmopolitan  
43Grey heronArdea cinerea group6134As for (b)6134
44Stilt****Recurvirostridae5387Recurvirostridae, Ibidorrhynchidae, Haematopodidae60121
45Ringed ploverCharadrius hiaticula group s. str.5265Charadrius hiaticula group s. lat.6679
46LapwingNot cosmopolitan  Not cosmopolitan  
47Kentish ploverNot cosmopolitan  Not cosmopolitan  
48Tern**Core Sterna51313Core Sterna, Chlidonias63317
49GullCore Larus51931Larus s. lat.61245
50CurlewNot cosmopolitan  Not cosmopolitan  
51SnipeGallinago gallinago superspecies5122 Gallinago 64716
52Stint****Core Calidris s. str.5006Core Calidris s. lat.61110
53Common sandpiper Actitis 7012As for (b)  
54Greenshank****Tringa nebularia group5552Not cosmopolitan  
55RedshankTringa totanus group6255As for (b)  
56Green sandpiper**** Tringa ochropus 5012Not cosmopolitan  
57OspreyPandionidae5711Not cosmopolitan  
58Golden eagleCore Aquila, Hieraeetus51815Aquila, Hieraeetus, Oroaetus, Spizaetus, Stephanoaetus60229
59Kite**Milvinae, Haliaeetinae54116Milvinae, Haliaeetinae, Butastur, Bursarellus, Geranospiza65826
60BuzzardCore Buteo54013 Buteo, Leucopternis 60332
61HarrierCore Circus61813As for (b)61813
62Hawk*Accipiter s. str., Astur, Cooperastur54115Accipiter s. str., Astur, Cooperastur, Leucospiza, Urospiza, Paraspizias65833
63Barn owl**** Tyto 51315Not cosmopolitan  
64Little owlSurniini, Athenini54334Not cosmopolitan  
65Tawny owlAsionini5268Not cosmopolitan  
66Eagle owlBubonini52821Bubonini, Strigini64242
67Kingfisher****Cerylidae, Alcionidae58270Cerylidae, Alcionidae, Alcedinidae68495
68Spotted woodpeckerDendropicini587109Core Picinae616177
69Green woodpecker*Picini, Malarpicini54561Included in above  
70FalconFalco subbuteo group, F. columbarius61611As for (b)61611
71Peregrine Falco peregrinus 5591Falco peregrinus and relatives6345
72Crow*Core Corvus59739Core Corvidae691103
73Wren****Polioptilidae, Certhiidae, Sittidae, Troglodytidae502124Not cosmopolitan  
74StarlingMimidae, Sturnidae, Rhabdornitidae611145As for (b)611145
75ThrushCore Turdus60762As for (b)60762
76BuntingEmberizidae562145Not cosmopolitan  
77ChaffinchFringillidae504135Not cosmopolitan  
78Pipit Anthus 61340As for (b)61340
79LarkAlaudidae51393Not cosmopolitan  
80Sedge warbler****Acrocephalidae, Megaluridae, Bernieridae, Donacobiidae517109Not cosmopolitan  
81Willow warbler****Phylloscopidae, Sylviidae, Zosteropidae, Paradoxornithidae, Timaliidae, Aegithalidae, Cettiidae533520Not cosmopolitan  
82Barn swallow Hirundo rustica 6141As for (b)6141
83Red-rumped swallow* Cecropis, Petrochelidon 57814Not cosmopolitan  
Figure 1.

Phylogeny of the tetrapod vertebrates indicating the least inclusive lineages with cosmopolitan terrestrial distributions, according to at least one of three criteria (a–c) (light grey, one criterion; dark grey, two criteria; black, all three criteria). Criterion (a), present in at least 215 of 281 global equal-area cells that contain at least 10% land; criterion (b), present in at least 500, or for criterion (c) at least 600, of 821 global ecoregions delimited by Olson et al. (2001). Humans and species that attained cosmopolitan status as a result of human agency were excluded. Numbers correspond to those in Table 1, where names are provided. Reference details are provided in Appendix S1.

Among these lineages, there were clear cases of both environmental and historical limitations to colonizing the unoccupied parts of the world [biomes and regions, respectively, in Fig. 2, based on criterion (c)]. Cosmopolitan lineages, according to both criterion (b) and criterion (c), were few in number in tundra and island ecoregions but numerous elsewhere, and most numerous in southern China and northern Indochina, where four ecoregions had representatives of all the cosmopolitan lineages listed here. The number of cosmopolitan lineages was also high in the savanna and temperate forest ecoregions of the Old World, and comparatively lower in tropical rain forest and in the New World. In agreement with predictions from dispersal theory, there was indeed a paucity of cosmopolitan lineages in historically isolated regions such as Madagascar and Australia [30–40 out of 77 lineages according to criterion (b) through much of Australia; Fig. 3a,b].

Figure 2.

Maps exemplifying historical and environmental limits to the natural distributions of cosmopolitan lineages (present in at least 600 of 821 global ecoregions delimited by Olson et al., 2001). (a) All regions except some island regions (mouse lineage); (b) all regions except Australia and island regions (deer lineage); (c) all biomes except cold and cool northern biomes (chameleon lineage); (d) all biomes except tundra and desert (green woodpecker lineage; also regionally absent in Australia and some island regions); and (e) all biomes except rain forest (lark lineage; also regionally absent in South America). Lineages were defined as the minimum-spanning clades that meet the 500-ecoregion criterion for being considered cosmopolitan (see text); as such, lineages are not strictly equivalent to the meaning of the corresponding English common name (e.g. chameleon lineage also includes Agamidae and Iguanidae; mouse lineage excludes some African and Madagascan representatives of Muridae s. lat.).

Figure 3.

Patterns in cosmopolitan lineage diversity and origination for terrestrial tetrapod vertebrates globally: (a, b) global patterns in the number of cosmopolitan lineages present; (c, d) the percentage of all species present belonging to cosmopolitan lineages; (e, f) the number of lineages present that are sister to cosmopolitan lineages; and (g, h) the diversity of range-restricted lineages that are sister to cosmopolitan lineages (see Materials and Methods for details). The 500-ecoregion criterion for considering a lineage cosmopolitan was employed in panels (a), (c), (e) and (g), and the 600-ecoregion criterion in panels (b), (d), (f) and (h). The six colour shades, from yellow (through green) to blue, represent the six categories in a typical box-and-whisker plot (i.e. top outliers, top whiskers, top quartile to median, median to bottom quartile, bottom whiskers, and bottom outliers, respectively).

The percentage of species belonging to cosmopolitan lineages was variable on islands (with both very high and very low values, more often high in ecoregions comprising young isolated islands) and in arctic regions, high in desert and grassland regions (both in the Old World and in North America) and in some rain forest areas [e.g. 29.3% based on criterion (c) in the Borneo montane rain forest ecoregion], but low in Australia and South America [lowest mainland values: 11.8% according to criterion (b) in the Purus–Madeira moist forest ecoregion, and 7.9% in the Gibson Desert ecoregion according to criterion (c)]. High values were noted in cold northern ecoregions and in Southeast Asia, especially when employing criterion (b) (cf. Fig. 3c,d).

Lineages sister to cosmopolitan lineages were also concentrated in the Old World [using criterion (b); maximum value in the South China–Vietnam subtropical evergreen forest ecoregion, where there are representatives of 44 of the 77 lineages recognized based on this criterion], but were also present in the Americas, and proportionally more numerous there when employing criterion (c) (Fig. 3e,f). However, in the case of criterion (c), more of the sister lineages are cosmopolitan themselves, or are at least widespread. Low values according to both criteria were noted in arctic and island regions, and in Australia. When focusing on range-restricted lineages, the unweighted map tended to overemphasize ecoregions that have single-species narrow endemic sister lineages, while the log-transformation evened out values to the point where patterns were lost (data not presented). We present here the square-root transformation for weighted values; the absolute numbers of range-restricted lineages showed a similar pattern (not presented). In the square-root-transformed data, the areas with most range-restricted sister lineages were located in the savanna regions of Africa (Fig. 3g,h).

When employing criterion (c) (Fig. 3h), the greatest proportional contribution was made by the Kurrichane thrush (Turdus libonyanus, the senior partner in a two-species lineage sister to the cosmopolitan thrush lineage), the giraffe family (Giraffidae, probably sister to the cosmopolitan deer lineage, which also incorporates Bovidae), the honey badger (Mellivora capensis, sister to the weasel–otter lineage) and the Nesomyinae + Petromyscinae rodents (sister to the cosmopolitan mouse lineage), all of which are African endemic lineages or nearly so. When employing criterion (b), there was also a high representation of such lineages in East and Southeast Asia (Fig. 3g), and lineages (nearly) endemic to this area that contributed substantially to weighted values included wren-babblers (Pnoepyga, sister to the cosmopolitan sedge warbler lineage), the ibisbill (Ibidorhyncha struthersii, sister to the cosmopolitan stilt lineage), the Baikal teal (Anas formosa, sister to the cosmopolitan shoveller lineage), a basal squirrel lineage (Ratufa, together with Sciurillus, sister to all other squirrels), and a snake lineage (Ahaetulla + Dendrelaphis Chrysopelea, sister to the smooth snake lineage). African lineages that contributed substantially to these values under criterion (b) were mostly distinct from those under criterion (c), including only the giraffe lineage in common, with the other lineages being the African black duck (Anas sparsa, sister to the mallard lineage), the white-faced owls (Ptilopsis, sister to tawny owls), cordyloid lizards (sister to the skink lineage) and the longclaws (MacronyxTmetothylacus, sister to the pipit lineage) (see references in Appendix S1). With criterion (c), values in East and Southeast Asia are lower, and values comparable to these are encountered in South America (Fig. 3h).

Cosmopolitan lineages with higher body mass and the ability to fly were represented by significantly fewer species (ANOVA and t-test respectively, P < 0.001 in both cases; see Fig. 4 for a visual representation of the differences).

Figure 4.

The influence of (a) flight ability and (b) body size on the number of terrestrial tetrapod vertebrate species required to make a lineage cosmopolitan (present in at least 500 of 821 global ecoregions delimited by Olson et al., 2001).

Discussion

Caveats

The WildFinder data set used here (World Wide Fund for Nature, 2010) is the most comprehensive data set available for this scale, and we would argue that, for a first assessment of such a broad topic, point data (available for more than half the groups analysed here) are not necessary. The WildFinder data set is not entirely complete, and reflects not only natural patterns but also human-caused extirpation, although not introductions (some of these aspects are reflected in the maps presented in Fig. 2). After careful examination of these effects, we concluded that false absences due to incomplete data are unlikely to affect our overall patterns. Human-driven changes have, however, probably resulted in several naturally cosmopolitan lineages not being cosmopolitan any more. For example, prior to human expansion and hunting there were probably cosmopolitan elephant and camel lineages, although one can also argue that, from that period, there were also substantial natural climate-driven range changes in other lineages (see, for example, Brook & Bowman, 2004). Climate-driven range changes, together with continental arrangement, also drove distributional patterns and the number of cosmopolitan lineages through geological time (for some interesting deep-time patterns, see Ezcurra, 2010).

Although the understanding of phylogenetic relationships in tetrapod vertebrates has improved dramatically in recent years, some aspects remain unclear, and this may affect the reliability of our results. For example, the lack of a fully resolved phylogenetic tree for falcons prevented us from gaining a good understanding of relationships in this group, and, while we list two lineages in this group as cosmopolitan, the existence of a third lineage cannot be ruled out. Furthermore, recent phylogenetic treatments of passerine birds (Irestedt et al., 2011) revealed the existence of two borderline cosmopolitan lineages in Sylvioidea that would not have been suspected 10 years ago [the sedge warbler lineage only just meets criterion (b) owing to the inclusion of the Californian genus Chamaea, and likewise the willow warbler lineage owing to the South American Donacobius]. Interestingly, in both of these cases, the inclusion of these genera also means that these lineages occur in the New World, which they otherwise would not (and all other 81 lineages in Table 1 did). We are aware that a criterion based on broader geographical units could add further cosmopolitan lineages (such as pelicans, storks or cranes, found on nearly all continents, although not necessarily widespread within continents). However, the New World representation of all lineages on our list suggests that such an alternative approach would not exclude any of the lineages already identified here.

While we doubt that numerous additional cosmopolitan lineages will emerge as a result of new phylogenetic treatments or increased resolution, we do, however, acknowledge that changes in the understanding of phylogenetic relationships are likely to affect the identity of some sister groups. It is worth noting that the giraffe lineage, listed here as sister to the deer lineage, notwithstanding possible changes in the phylogeny (Spaulding et al., 2009), also previously occurred in Eurasia and may have originated there (Prothero & Foss, 2007). The reliability of our results in the face of changed sister lineages is to some extent tested by including analyses based on both criterion (b) and criterion (c) (i.e. by increasing the cut-off value from 500 to 600 ecoregions). The results are only partly encouraging, as patterns do change somewhat (high values in Africa are noted in both cases, but not in Southeast Asia).

Vicariance versus dispersal

A quantitative analysis of the ages of cosmopolitan lineages will only be possible once dated phylogenies are available for all groups. Based on those available at present, vicariance-driven cosmopolitanism can be fully supported in only one of the 83 lineages (the oldest split within any one of these lineages, the turtle lineage, being dated at c. 200 Ma, when most continents were still connected in Pangaea, and cosmopolitan status would be lost by all criteria if any sub-lineage was removed). In at least two other cases (just over 100 Ma for the chameleon and common lizard lineages), it is possible that, although too late for vicariance proper, the proximity between continents at the earliest within-lineage divergence time gave these lineages a dispersal headstart. For all other lineages, at less than 90 Ma, most continents were already separated by substantial sea barriers.

This does not preclude vicariance from being relevant at deeper nodes in broader tetrapod groups, for example placental mammals, core Anura and core Serpentes (Springer et al., 2003; Marjanović & Laurin, 2007; Pyron et al., 2011). These are groups of multiple cosmopolitan lineages, meaning that, although they may have originally diversified by vicariance, lineages within them later on colonized the world through dispersal from different starting points.

Different data would be necessary to discuss which dispersal paths were most important in facilitating the attainment of cosmopolitan status across lineages, although we have no reason to believe that these were fundamentally different from those illustrated by Donoghue (2008) for plants at various times during the Tertiary. For more than half the bird lineages, all one can say at this point is that they are more recent than the Cretaceous–Palaeogene boundary (see references in Appendix S1), although in fact some of the single-species lineages probably became cosmopolitan only in the last 1 Myr.

Out of where?

The largely distinct sets of sister lineages indicating the high diversity of range-restricted sister lineages in Africa when employing criteria (b) and (c) suggest that this pattern may be a genuine one, and that Africa is indeed an important area of origination for cosmopolitan lineages (‘out of Africa’ hypothesis; Waters & Roy, 2004; Kodandaramaiah & Wahlberg, 2007). The low diversity of sister lineages in the Americas is consistent with the New World's limited exchange of biota with the rest of the world (Procheş, 2006), although some sister lineages do occur only here (Icteridae sister to the bunting lineage; Falco mexicanus sister to the peregrine lineage; Perimyotis sister to the barbastelle lineage; Romerolagus sister to the hare lineage; for other taxa, also see a noctuid moth lineage in Kergoat et al., 2012). Low sister-lineage diversity on islands is consistent with lower colonization rates from islands to mainland (Bellemain & Ricklefs, 2008).

Our results, moderately reliable though they are, support the ‘out of Africa’ and ‘out of Southeast Asia’ hypotheses. These are indeed the most biodiverse regions of the world besides tropical America, which is less likely to be the origin of global colonizations owing to longitudinal isolation (see Procheş, 2006). A different approach likely to provide the same results would be to model the likelihood of lineages dispersing to all other parts of the world from different starting points (cf. Ree & Smith, 2008).

Range and traits

Both flight and body size significantly influence the number of species necessary to make a lineage cosmopolitan, and there is no strong interaction between the two variables, with flying lineages being evenly distributed across body size and, if anything, under-represented in the largest size class, which is dominated by mammals (see raw data in Appendix S1). However, within birds, body size seems also to be important, which is possibly related to the ability to sustain flight over long distances (Gaston & Blackburn, 1996). Bird species diversity is almost equally divided between passerines and non-passerines, but there are 47 cosmopolitan lineages among non-passerines and only 12 among passerines (Table 1).

Another factor likely to be important here is long-range migration. The existence of distinct summer and winter ranges in numerous bird lineages (all counted as presences in the WildFinder data set) greatly increases the total range size. Apart from this methodological issue though, migration enhances the likelihood of breeding range expansion. Historically, migration was linked to fluctuations in resource availability. In most lineages, migration is likely to have arisen in relation to the Tertiary cooling, although in some cases it may be older (Mummenhoff & Franzke, 2007). The existence of migration routes between Africa and Eurasia may actually be responsible for the over-representation of African lineages among lineages sister to cosmopolitan ones. Bird lineages including migratory species, such as the barn swallow and pipit lineages, may support this hypothesis, but dated trees are needed to confirm the timing of their dispersal out of Africa.

The way forwards

While our study was limited to vertebrate data, patterns in other groups of terrestrial organisms are likely to be similar, in the sense that multiple lineages will occur in groups with good dispersal abilities. It would be interesting to see how better dispersal opportunities in the marine environment affect the number and characteristics of marine cosmopolitan lineages, and also to link secondary marine transitions (Procheş, 2001) with global range-size success, as compared with fully terrestrial relatives.

When looking at invertebrates and plants, however, the body size relationship is likely to change. Very small animals and wind-dispersed plants (small propagule size in both cases) are likely to be over-represented (Fenchel & Finlay, 2004), thus making the body size–species per lineage relationship a U-shaped one. Another notable difference is the almost complete absence among invertebrates of seasonally long-range migratory species, partly owing to rapid generational turnover (Holloway & Jardine, 1968), with the exception of a few groups of Lepidoptera (Holloway & Nielsen, 1999).

Among both invertebrates and plants, cosmopolitan distributions are probably most common among r-strategists, which, from a human perspective, is relevant to invasiveness and economic impact. A large proportion of weedy plants belong to a small number of widely distributed lineages (Lososová et al., 2006), and one cosmopolitan clade in noctuid moths is specifically named the ‘pest clade’ (Holloway & Nielsen, 1999). The distinction between cosmopolitan species and multiple-species cosmopolitan radiations is particularly relevant from an invasion perspective. Single-species cosmopolitan lineages are generally exceptional dispersers that are already part of the indigenous species lists in most areas, and at some point would probably reach most parts of the world where they are currently absent. Multi-species lineages, however, may be less likely to be able to take these last few steps in the short term, and their introduction to regions where they are not naturally present (often historically isolated) would present the risk of impacting biodiversity and ecological processes heavily, considering that they may fill niches that are as yet vacant (Denslow, 2003; Blackburn et al., 2004). In this context, a trait-centred approach to understanding the success of cosmopolitan lineages and subsequent trait-mapping on phylogenies could be the next big step in the fast-growing field of invasion phylogenetics (Procheş et al., 2008).

Acknowledgements

Sandun Perera helped with lineage identification, and Timothy Wiggill helped with GIS mapping. Jeremy Holloway, who commented on two earlier versions of the paper, provided insect examples and stressed the importance of bird migration. We also thank Martin Ezcurra and one anonymous referee for useful suggestions. Financial support was provided by the National Research Foundation of South Africa (Incentive Funding for Rated Researchers to Ş.P.) and UKZN (Post-doctoral Fellowship to S.R., 2011).

Biosketches

Şerban Procheş is an associate professor of biogeography at the University of KwaZulu-Natal, with research interests including global regionalization, as well as the distributional patterns of ancient lineages and recent radiations.

Syd Ramdhani is a lecturer in the School of Life Sciences at the above institution. His research interests include systematics, evolution, biogeography and ecology.

Ancillary