Increasing attention is being focused on biomes, their evolution and assembly (Pennington et al., 2004). Biomes are broad biogeographical regions defined by climate, vegetation structure and ecophysiology. Narrowly, a ‘biome’ is a recognizable area of endemism containing a ‘biota’. In one viewpoint, biomes are analogous to organismal lineages, responding as a whole to events in earth history and, in the extreme view, classifiable like organisms. Thus a biota that was widespread in the Mesozoic of Gondwana might have fragmented and differentiated into new biomes by vicariance as the supercontinent split apart. A contrasting view is that biomes have accreted stochastically from individual taxa, each on its own trajectory through time and space. In this view, biome composition is dynamic because taxa respond differently to environmental changes (by adaptation, radiation and extinction) and opportunities to migrate to new areas. A biome might contain any combination of both kinds of elements. For two decades, the first scenario has been the province of cladistic biogeography; the second of ecological biogeography. Nevertheless, a phylogenetic framework is needed to test hypotheses about events occurring over macroevolutionary time, irrespective of whether the events were driven by vicariance or ecological factors.
To differentiate between these scenarios, comparative studies are needed using representative samples of taxa in biomes. Such studies are few (e.g. Sanmartín & Ronquist, 2004), and even fewer have employed molecular dating to test the predicted timing of divergences (Givnish & Renner, 2004; Pennington et al., 2004). Specific questions relevant to biome assembly are: (1) what is the geographical pattern of sister taxa among biomes? (2) is there a prediction of congruent pattern resulting from historical events (such as vicariance or opening of a migration corridor)? (3) if migration is inferred, what was its direction between biomes? and (4) did the timing of divergence events coincide in taxa for which congruence is inferred?
These questions require inference of the ancestral area of the most recent common ancestor of sister taxa in different biomes. The direction of migration is important because only migrations into a biome are relevant to its assembly. Also, knowledge of the sources of immigrating taxa is needed to test whether the biome evolved as a unit.
No historical biogeographical method adequately infers ancestral areas or direction of migration. If the area occupied by an ancestor were known, then the direction of dispersal (if any) into the areas occupied by its descendants could be deduced. Cladistic biogeography implicitly ignores dispersal and infers ancestral areas by assuming ancestral cosmopolitanism. Other methods use variants of parsimony or maximum likelihood to infer ancestral areas from extant areas analogously to inferring ancestral traits. For example, a standard parsimony approach reasons that, if a taxon restricted to area A is nested within a paraphyletic group of taxa that are all restricted to area B, then B is inferred as ancestral followed by migration from B to A. Although this method is widely used, it has several problems. Numerous taxa are known as fossils in areas where they are no longer extant (Hill, 2004), but current methods model extant distributions only, excluding a priori the possibility of an ancestral range outside the extant range. Also, the analogy between areas and traits is not strictly applicable, because area ‘inheritance’ depends on the mode of speciation (Cook & Crisp, 2005). Third, the method fails to take account of processes that could have caused multiple dispersals in the opposite direction (from A to B), such as prevailing wind and ocean currents, leading to a paraphyletic area B (Cook & Crisp, 2005).
A congruent pattern of sister-group relationships between taxa in different biomes might be predicted if these biota share a common geographical history (such as vicariance or mass migration). This suggests a null hypothesis that taxa have independent patterns in time and space. In contrast, cladistic biogeography operates under a null hypothesis of congruence because it assumes vicariance is the dominant process. However, vicariance should not be assumed as additional processes, such as immigration and extinction, are known to influence organismal distributions. Therefore congruence is a hypothesis that requires testing against a null hypothesis of independence among taxa.
Congruence due to a common environmental cause is testable by dating divergences between lineages. For example, Poux et al. (2005) asked whether the sister-group relationships between Africa and Madagascar in four mammalian lineages could be explained by Gondwanan vicariance (> 84 Ma) or a land bridge (45–26 Ma). Using a relaxed molecular clock calibrated with several fossils, they rejected Gondwanan vicariance in all four cases: the oldest 95% confidence limit for any divergence (between Malagascy lemurs and Lorisiformes) was 69.6 Ma. Although the land bridge could not be rejected because all four sets of confidence limits overlapped the hypothesized period, synchrony of the four divergences was rejected.
In a recent issue of this journal, Galley & Linder (2006) took a similar approach to the history of the Cape Floristic Region (CFR). This is a mega-diverse biome, one of seven floral kingdoms in the world. Its huge diversity is concentrated into a small area, yet 70% of species are endemic despite a broad land link to the rest of Africa. Ecologically and floristically, it is sharply distinguished from adjacent biomes. Galley & Linder compared 13 dated phylogenies representing the dominant components of the heathy fynbos vegetation that characterizes the CFR (e.g. Erica, Proteaceae, Bruniaceae, Restionaceae and Phylica). CFR taxa have sister groups in Australia, New Caledonia, New Zealand and South America (the so-called Gondwanan track), tropical Africa and Europe. Galley & Linder (2006) asked whether these links are ancient and vicariant or the results of more recent migration. They also asked whether the divergences between CFR taxa and those elsewhere were consistent with a common driving event such as vicariance or concerted dispersal.
Galley & Linder found no congruent pattern of area relationships. Many more temperate taxa had links to Australasia than to South America, which they reported as surprising because the south Atlantic is narrower (6000 km) than the Indian Ocean (8000 km). However, Galley & Linder did not consider West Wind Drift, which would favour dispersal across the southern Indian Ocean over that across the South Atlantic because West Wind Drift is stronger in the former.
The divergence times found by Galley & Linder appeared too young to support a hypothesis of Gondwanan vicariance. The southern Indian Ocean probably opened between 120 and 100 Ma, earlier than any of the mean divergence times found in their study. Even the possibility that the Kerguelen Plateau provided a stepping stone between CFR and Australia until c. 90 Ma could account for the divergence in only four dated divergences between CFR and Australasia. Similarly, the single dated CFR–South American divergence appeared too young to be the result of the opening of the South Atlantic c. 120 Ma. However, in the absence of confidence limits on some of these estimates, Gondwanan vicariance cannot yet be completely ruled out.
Using the parsimony criterion, only four of the phylogenies examined by Galley & Linder indicated a directionality of migration between CFR and other areas. Two were inferred to have come from Australia (Ehrharta and Restionaceae); one from elsewhere in Africa (Danthonioideae); and one from other sources in the southern Hemisphere (Iridaceae, in part). On the other hand, migrations of taxa from the CFR to the high mountains of tropical Africa were inferred in eight different taxa, rejecting previous hypotheses of vicariance or migration in the opposite direction. An obvious test of this hypothesis – whether the age of the tropical African mountains is consistent with the timing of the inferred migrations – was not carried out.
Overall, the analysis of Galley & Linder (2006) hammers another nail into the coffin of the Gondwanan dogma of a vicariant temperate southern hemisphere flora (McGlone, 2005). Their study provides strong evidence that the CFR has not evolved as a single biotic unit through space and time. Rather, it seems to have been assembled stochastically over a prolonged period (at least the last 90 Myr) with different lineages immigrating from (or emigrating to) external areas in several directions. Similar results have been found by other comparative studies (Givnish & Renner, 2004; Sanmartín & Ronquist, 2004; Poux et al., 2005). Further evidence against the unitary view of biomes is provided by ‘no-analogue’ palaeocommunities (Ackerly, 2003; Hill, 2004), which are thought to have combined environmental parameters and taxa that do not co-occur today. Environmental tracking by taxa that once co-existed in these communities has apparently led to their descendants occupying different areas, and thus the biomes themselves have changed. Application of ecological principles to historical biogeography has a controversial past, but there is an increasing view that reintegration of ecology and biogeography is necessary for the latter to progress (Ackerly, 2003; Cook & Crisp, 2005; Riddle, 2005).