The Cape of southern Africa and the Mediterranean Basin, two of the world's five mediterranean-climate biodiversity hotspots, are exceptionally species-rich and constitute a well-described example of ecological convergence. However, the area-adjusted plant species diversity of the Cape is on average more than double that of the Mediterranean Basin. Here, we investigate the causes of this diversity asymmetry by drawing on phylogenetic information from a variety of plant groups and focusing on three competing hypotheses: diversity disparities arising from differential clade ages, diversification rates or diversity limits.
Cape of southern Africa and the Mediterranean Basin.
We reviewed a variety of studies in order to contrast the geography, geomorphic history, biogeographical connectivity and ecological context in the two hotspots. We also tested the relationship between clade age and species richness in both regions based on phylogenetic information from 39 clades.
Clades are on average older in the Cape than in the Mediterranean Basin. Clade age is a strong predictor of species diversity in the Cape, suggesting that diversity-dependent regulatory mechanisms may be weak. In contrast, we failed to find a relationship between age and diversity in the Mediterranean Basin, indicating that a diversity limit may have been achieved in multiple clades.
The Cape has a higher species density than the Mediterranean Basin owing to a combination of older clade ages, high rates of diversification in certain lineages and an exceptionally high upper limit to diversity. High richness in the Cape is linked to long-term lineage persistence in a heterogeneous but stable evolutionary context. In contrast, the climatically unstable Mediterranean Basin has offered fewer opportunities for diversity accumulation in the long term (owing to high extinction rates), but appears to be a hotspot of recent rapid speciation.
Diversity asymmetries among regions and clades have received renewed attention in recent years, fuelled by new evidence from phylogenies and the fossil record that highlights the need to incorporate evolutionary history and ecology in order to explain global diversity patterns (Rabosky, 2009; Kisel et al., 2011; Wiens, 2011; Etienne et al., 2012). Many of the ideas proposed by recent studies on species diversification remain controversial, particularly with respect to the main predictors of spatial and taxonomic imbalances in species richness (Hardy & Cook, 2012; Cornell, 2013). Some authors advocate a key role of differential clade ages and/or diversification rates in generating disparities in species richness between regions (Wiens, 2011). Several studies, however, have proposed that the diversity of a region is regulated mostly by ecological diversity-dependent processes (Rabosky, 2009; Etienne & Haegeman, 2012). The relative importance of clade ages, diversification rates and ecological limits in determining species richness has not yet been simultaneously examined in regions that share similar environments but that exhibit striking differences in species richness. In particular, we still know very little regarding why regions at equivalent latitudes and climates often support very different numbers of species (Schluter & Ricklefs, 1993; Valente et al., 2011).
The mediterranean-climate regions of the globe (all located between 32° and 44° in both the Northern Hemisphere and the Southern Hemisphere) are very interesting systems with which to address these questions, as they exhibit disparities in species diversity patterns despite their well-described convergence in ecological conditions (Cowling et al., 1996). Mediterranean-type ecosystems harbour an exceptional level of plant species diversity: they occupy only 5% of the world's land surface but contain over 48,000 species of vascular plants, nearly 20% of the world's total (Dallman, 1998). All five mediterranean-climate floristic regions of the globe (California, central Chile, south-western Australia, the south-western Cape of southern Africa, and the Mediterranean Basin) have been included in the list of biodiversity hotspots (Mittermeier et al., 2004), and the causes of their high diversity have been the subject of much debate (Linder, 2003; Comes, 2004; Hopper & Gioia, 2004; Cowling et al., 2009; Kadereit & Baldwin, 2012). There is still much uncertainty, however, regarding how the processes of species genesis in mediterranean floras took place and whether they have been homogeneous throughout the five regions.
In this review, we investigate evolutionary patterns in mediterranean-climate regions by focusing on plant groups that occur within two of these biomes: the Cape of southern Africa and the Mediterranean Basin (Fig. 1). We adhere to the circumscription of the Cape Floristic Region (hereafter CFR) as defined by Goldblatt & Manning (2002), and of the Mediterranean Basin (hereafter MB) as defined by Quézel (1985), as these definitions encompass only the areas with a strict mediterranean climate (a wider definition of the CFR region including semi-arid regions has been proposed by Born et al., 2007).
The CFR and the MB are situated at equivalent latitudes and are characterized by a seasonal climatic regime with summer drought (very low rainfall for 3–4 months) and winter rainfall (at least 65% of precipitation occurs in winter), which has led to a striking convergence of ecological conditions (Cody & Mooney, 1978). Both territories have high habitat heterogeneity and are dominated by evergreen, sclerophyllous vegetation adapted to seasonal drought (Mooney & Dunn, 1969). Both are strongly influenced by a frequent fire regime, which plays a prominent role in ecosystem and evolutionary dynamics (Verdú et al., 2009; Keeley et al., 2012). The two regions display a high percentage of endemic species and high diversities at alpha, beta and gamma levels (Goldblatt & Manning, 2002; Thompson, 2005).
Despite important similarities in ecological factors and community structure (Cody & Mooney, 1978), the CFR and the MB differ greatly in various diversity parameters. The MB has the highest total plant diversity of all mediterranean biomes, harbouring c. 25,000 species, 50% of which are endemic (Médail & Quézel, 1997). In contrast, the CFR is by far the richest mediterranean-climate region for its land area, with at least 9000 plant species, 69% of which are endemic (Goldblatt & Manning, 2002; Goldblatt et al., 2005). Importantly, overall, areas in the CFR support over twice as many species as similar-sized areas in the MB (Cowling et al., 1996). Indeed, the CFR contains double the number of species predicted by global models of biodiversity and is thus considered one of the most species-rich regions on Earth (Kreft & Jetz, 2007; Kier et al., 2009).
Here, we investigate the differences in diversification patterns between the CFR and the MB by drawing on data from geological, ecological, phylogenetic and phylogeographical studies. We place particular emphasis on clades presenting a CFR–MB disjunction (Tables 1 & 2), as these groups offer the best potential to disentangle drivers of diversification. We aim to: (1) compare and contrast the evolutionary context in the two regions; (2) test the relationship between clade age and diversity in both biomes; and (3) understand the net contribution of three hypotheses, namely differential clade ages, diversification rates, and spatio-ecological limits to diversity, in accounting for the differences in diversity between the two regions.
Table 1. Examples of plant genera that occur in both the Mediterranean Basin (MB) and the Cape Floristic Region (CFR). Data from Tutin (1993); Mabberley (1997); Goldblatt & Manning (2000, 2002); Greuter et al. (2007). Genera that have been phylogenetically studied to date are marked in bold (see Table 2 for details)
No. of species world-wide
No. of CFR species
No. of MB species
Table 2. Plant clades presenting a Cape Floristic Region–Mediterranean Basin (CFR–MB) disjunction for which phylogenetic, dating and evolutionary hypotheses are available
Ancient age and moderate rates of diversification in the CFR, where one-third of speciation events are associated with pollinator shifts; remaining speciation events unexplained. Recent arrival and low rates of diversification in the MB.
Evolutionary Context in The Cape and The Mediterranean Basin
In order to understand how diversification processes may have varied between the two regions it is first essential to identify biome differences in physical, historical and ecological factors. The major differences between the CFR and the MB are described in the following sections.
Area, geography and diversity patterns
Spanning three continents, the MB is the largest mediterranean-climate biome in terms of area (2,300,000 km2, Fig. 1), being 25 times larger than the CFR (90,000 km2), the smallest of the five regions. At large spatial scales (10–106 km2), the slope of species–area relationships (SARs) for mediterranean regions is homogenous (c. 0.2 for a power model; Kier et al., 2005), and therefore the intercepts can be directly compared. At this scale, the CFR clearly displays a higher density of species, as shown by the fact that the intercept of the SAR for this region is 2.2 times higher than that for the MB (Cowling et al., 1996). The high diversity of the CFR was clearly evidenced when we fitted a power-model SAR to the five mediterranean-climate regions (Fig. 2). Although the diversity of both the CFR and the MB falls above the SAR, the CFR is over twice as diverse as expected for its area (Fig. 2). As an illustration of this, the mediterranean part of continental Greece (9 × 104 km2; Médail & Quézel, 1997) is of roughly the same area as the CFR, but harbours fewer than half as many plant species (c. 4000 species vs. c. 9000; Médail & Quézel, 1997).
The CFR is also more diverse than the MB at the point scale (1 m2; Keeley & Fotheringham, 2003). However, at the 1000-m2 community scale, the shrublands and woodlands in the eastern MB were in fact more diverse than both CFR fynbos and renosterveld communities (Keeley & Fotheringham, 2003). The fact that the CFR presents a much higher diversity at regional and point scales but often a lower diversity at intermediate scales when compared with the MB suggests that beta diversity is a major contributor to the overall plant species richness in the CFR. Indeed, several studies seem to support the fact that species turnover between habitats is exceptionally high in the CFR (Simmons & Cowling, 1996; Linder, 2003).
Overall, the MB is more geographically heterogeneous than the CFR, as it is typified by sea barriers, straits, islands and archipelagos (Rosenbaum et al., 2002), all of which are lacking in the CFR. The Mediterranean Sea, the largest inland sea in the world, is the most significant geographical barrier for plant dispersal (Escudero et al., 2008b), but the region is also dotted with important mountain chains, reaching elevations over 2500 m (e.g. Atlas, Betic, Olympus, Taurus). A pattern of high plant endemism associated with geographical barriers is strongly manifested within the circum-Mediterranean region (Thompson, 2005). Several of the most species-rich MB clades display a high level of taxonomic turnover across different archipelagos, islands, peninsulas and mountains, often with little ecological differentiation; for example Anthemis (Lo Presti & Oberprieler, 2009), Antirrhinum (Vargas et al., 2009), Centaurea (Font et al., 2009), Cistus (Guzmán et al., 2009), Dianthus (Balao et al., 2010) and Nigella (Bittkau & Comes, 2009). These patterns suggest that diversity has been strongly driven by allopatric speciation in this region.
With no sea barriers, the CFR region is geographically more homogeneous and compact. The major barrier to biological movements in this territory is the Cape Folded Belt, a series of moderately elevated mountains that delimits coastal, intermontane and inland plains (Cowling et al., 2009). Certain clades display high numbers of single-mountain endemics, such as Iridaceae (Goldblatt & Manning, 1998) and Proteaceae (Rebelo, 2001), implying a role of geographical isolation in speciation at least in these groups. However, the best predictors of reproductive isolation in the CF seem to be ecological rather than geographical (see ‘Ecological setting’ below).
In short, the abundant sea and orographical barriers in the large MB have promoted allopatric speciation, whereas small areas have been the arena for speciation in the CFR.
Geomorphic and climatic history
Since the Oligocene, the history of the MB has been more unstable and complex than that of the CFR in terms of both geomorphic and climatic events (Thompson, 2005; Cowling et al., 2009). Situated at the convergence of the African, Arabian, Eurasian and Iberian tectonic plates, the circum-Mediterranean region has undergone constant and dramatic geomorphic mutation. The late Oligocene and the Miocene were periods of important horizontal and vertical movements of land masses, with the break-up of continents, orogenesis, and volcanic activity in both the east and west of the MB (Robertson, 1998; Rosenbaum et al., 2002). In the late Miocene, the closing of marine passages between the Atlantic Ocean and the Mediterranean Sea led to increasing desiccation, culminating in the Messinian salinity crisis (5.3–5.9 million years ago, Ma; Duggen et al., 2003; Krijgsman et al., 2010). Subsequently, the Pliocene and Pleistocene were periods of island formation and intense orogenesis (particularly in the central MB; Rosenbaum et al., 2002). This complex history has most certainly led to high rates of isolation and speciation, but also to extinction, and possibly to the repeated extirpation of floras in areas of similar size to the CFR.
In contrast, the late Cenozoic in the CFR has been characterized by a comparatively stable geomorphic evolution. The region had a period of tectonic stability in the Oligocene, which was followed by cycles of orogenic uplift in the Miocene and Pliocene associated with the Post-African I and II erosion cycles (Partridge & Maud, 2000; Cowling et al., 2009). In the Quaternary there have been no major geomorphic phenomena in the CFR (Cowling et al., 2009), which suggests that rates of extinction may have been low.
The climatic histories of the two regions also differ. Most importantly, the onset of a true mediterranean climate took place earlier in the CFR. In southern Africa, progressive aridification of the climate began in the mid-Miocene (10–15 Ma), associated with the establishment of the Benguela upwelling system (Zachos et al., 2001; Cowling et al., 2009; Dupont et al., 2011). Approximately 10 Ma, the seasonal summer-dry climatic regime found in the present day in the CFR was established (Krammer et al., 2006; Roters & Henrich, 2010). The progressive aridification and the opening of new niches in the CFR in the last 10 million years (Myr) have been hypothesized to have triggered rapid speciation in several CFR clades (Verboom et al., 2009). Plant groups that have been shown to have diversified since the onset of the mediterranean climate in the CFR include Ehrharta (9.8–8.7 Ma; Verboom et al., 2003), Heliophileae (< 4 Ma; Mummenhoff et al., 2005) and Phylica (7–8 Ma; Richardson et al., 2001). However, not all species-rich CFR clades have experienced a speciation burst since the onset of aridification (see Appendix S1 in Supporting Information). Phylogenetic studies of clades such as Crotalarieae p.p. (46.3 Ma; Edwards & Hawkins, 2007), Pelargonium (30.5 Ma; Bakker et al., 2005) and Restionaceae (> 41–38 Ma; Linder et al., 2005) have shown that these clades had an ancient origin in this biome that clearly pre-dates the onset of the mediterranean-type climate.
In the MB, the profound climatic alterations that led to the modern climate took place much later than those in the CFR. Strong seasonality and aridification began by the mid–late Pliocene (c. 3.6 Ma), and only by the end of this period (c. 2.8 Ma) did the contemporary summer-drought regime stabilize (Suc, 1984; Suc & Popescu, 2005). These abrupt climatic changes caused the extinction of moisture-adapted lineages (Jiménez-Moreno et al., 2007), and have been shown to have coincided with diversification bursts of lineages that were pre-adapted to arid environments (Bell et al., 2012; Fiz-Palacios & Valcárcel, 2013), such as Antirrhinum (< 4.1 Ma; Vargas et al., 2009), the Cistus–Halimium complex (1.0 Ma; Guzmán et al., 2009), Dianthus (1.9–7 Ma; Valente et al., 2010a), Erodium (< 3 Ma; Fiz-Palacios et al., 2010) and Tragopogon (1.7–5.4 Ma; Bell et al., 2012). On the other hand, clades such as Cyclamen (29.20–30.10 Ma; Yesson et al., 2009), Linaria sect. Versicolores (6.45–16.53 Ma; Fernández-Mazuecos & Vargas, 2011), Narcissus (18.1–29.3 Ma; Santos-Gally et al., 2012) and Ruta (26.65–63.53 Ma; Salvo et al., 2010) are much older and seem to have survived the filtering processes associated with climatic change in the region.
The climatic history of the CFR and the MB has also differed significantly in the Quaternary. The severity of the temperature and moisture oscillations associated with Pleistocene glaciations was higher in the Northern Hemisphere than in the Southern Hemisphere (Markgraf et al., 1995; Jansson, 2003). Indeed, strong evidence suggests that long-term Quaternary climatic stability and summer-drought reliability were much higher in the CFR than in the MB (Cowling et al., 2004), which could have led to important differences in extinction and speciation rates. In the MB, clades such as Cedrus, Platanus and Syringa have been shown to have been severely affected by extinctions caused by Quaternary climate fluctuations (Postigo-Mijarra et al., 2010). In contrast, studies in the large CFR clades Gladiolus (Valente et al., 2011) and Pentaschistis (Galley et al., 2007a) have found evidence for low rates of extinction in the CFR in the Quaternary, in agreement with the hypothesis of high climatic stability for this region (Cowling et al., 2004; Galley et al., 2007a).
We conclude that the time-lag in the establishment of summer-drought climatic regimes in the CFR (10 Ma) compared with the MB (3.6 Ma), in combination with the more unstable climatic and geomorphic history in the MB, may have enabled more time for lineage accumulation with lower extinction rates in the CFR.
The MB is situated between three continents (Africa, Asia and Europe) and has thus been considered a ‘tension zone’ for plant lineages from different biogeographical origins (Comes, 2004). Several floristic regions are now known to have frequently acted as sources of species to the Mediterranean region (and vice versa), and its contemporary flora is rich in Irano-Turanian, Saharo-Arabian, Holarctic and subtropical African elements (Zohary, 1973; Quézel, 1978, 1985), with some clades often invading the region multiple times mostly from the east (Mansion et al., 2008) or from the south (Buerki et al., 2012). In contrast, the exceptional level of plant endemism in the CFR and its location at the southernmost tip of Africa have led to the suggestion that this continental region is ‘island-like’, with lower levels of connectivity than typical mainland areas (Kier et al., 2009). In several groups, migration out of the CFR into neighbouring summer-rainfall regions (namely the Drakensberg) has occurred more frequently than has migration into the CFR (Galley & Linder, 2006; Galley et al., 2007b; Valente et al., 2010b, 2011). Nevertheless, a recent study in Hyacinthaceae found that there have been a large number of independent colonizations of the CFR from sub-Saharan Africa (Buerki et al., 2012).
Plant movements have also taken place repeatedly between the two biomes and other mediterranean-climate regions (Burtt, 1971; Quézel, 1978). The MB has particularly strong affinities with California, with over 25 groups exhibiting a disjunct pattern between the two regions (Kadereit & Baldwin, 2012), whereas the CFR shares several groups with south-western Australia (e.g. Aizoaceae, Geraniaceae, Haemodoraceae and Proteaceae; Hopper & Gioia, 2004; Galley & Linder, 2006). Connections have also taken place several times between the CFR and the MB, leading to a considerable number of plant clades presenting an intriguing CFR–MB disjunct pattern (Tables 1 & 2). Possible land routes of connection between the two territories have been proposed, through the Sahara via the mountainous corridors of the central Mahgreb or near the Red Sea coast, and through tropical and summer-rainfall southern Africa via Afromontane arid tracks (Wickens, 1976; Quézel, 1978). In certain CFR–MB disjunct clades, however, long-distance dispersal mediated by birds has not been ruled out (e.g. Senecio; Coleman et al., 2003).
In summary, the MB presents a high level of connectivity with other regions, whereas the CFR appears to be comparatively more ‘island-like’, which has translated into its exceptional levels of endemism. Lineage exchange has occurred often between the two regions.
There is a significant turnover of species among numerous microhabitats in the MB (Naveh & Whittaker, 1979), suggesting that rapid local adaptation in association with allopatry may have been an important driver of speciation events. In the CFR, total microhabitat diversity may be lower overall, but some studies suggest that microhabitat heterogeneity within small areas may be unusually high (Simmons & Cowling, 1996). The two most common vegetation communities in the CFR – fynbos and renosterveld – are both typified by steep ecological gradients, particularly in terms of soil and hydrology (Linder, 2003; Silvertown et al., 2012). The fact that these gradients are associated with species turnover supports a role for ecological speciation processes in the CFR, at least for some of the largest clades, including Iridaceae, Orchidaceae, Proteaceae and Restionaceae (van der Niet & Johnson, 2009; Rymer et al., 2010).
One of the major hypotheses that has been put forward to explain the high species richness in the CFR highlights the role of edaphic specialization in generating diversity (Linder, 2003). Using phylogenetic sister-species comparisons, Schnitzler et al. (2011) identified soil type as the best predictor of speciation in Babiana, Moraea and Protea. However, it remains unclear why edaphic shifts should be more common here than in other regions (Barraclough, 2006). For instance, some areas in the MB, such as the southern Iberian Peninsula, display a similarly high diversity of soils, but harbour far fewer species than areas of similar size in the CFR (Ojeda et al., 2001).
The differences in angiosperm diversity between the CFR and the MB have also been hypothesized to be associated with regional differences in plant–pollinator interactions (Valente et al., 2012). One of the main hypotheses to explain the high plant diversity in the CFR proposes that a high percentage of speciation events within several lineages may have been driven by pollinator shifts (Johnson, 2010; Waterman et al., 2011; Valente et al., 2012). This hypothesis is supported by the fact that several of the species-rich CFR clades exhibit a high diversity of floral characters and pollination systems (Johnson, 1996). Examples of such clades include Babiana (Schnitzler et al., 2011), Disa (Johnson et al., 1998), Gladiolus (Valente et al., 2012) and Moraea (Goldblatt et al., 2002). In addition, it has been shown that a high percentage of species in southern Africa are typically pollinated by a single animal species (Johnson & Steiner, 2003). In contrast, the majority of plant species in the MB exhibit a generalist pollination strategy (Johnson & Steiner, 2000; Thompson, 2005), and large genera in the region possess a relatively low diversity of pollination systems (e.g. Astragalus, Antirrhinum). The lower diversity of specialized pollination systems in the MB could potentially be linked with the lower diversity of the specific pollinator clades that are associated with high floral specialization in the CFR (rather than with a lower total pollinator species density per se; Valente et al., 2012). For instance, the diversity of the most important pollinator group in both regions (long-tongue bees, Apidae) is higher in the CFR (Kuhlmann, 2009; Linder et al., 2010). In addition, several of the specific animal pollinator clades that are associated with high diversification in the CFR (Goldblatt & Manning, 2006) are completely absent in the MB (sunbirds, long-proboscid nemestrid and tabanid flies, hopliine beetles, among others).
Considered together, this evidence suggests that ecological speciation appears to have been more frequent in the CFR than in the MB, and that interactions with specific pollinator classes that are more prone to causing reproductive isolation in plants could have been more common in the former.
Relationship Between Clade Age and Diversity
We tested the relationship between clade age and species richness in the CFR and MB (Fig. 3), as this relationship can provide important insights into the roles of differential clade ages, diversification rates and ecological limits in explaining diversity asymmetries (Hardy & Cook, 2012; Rabosky, 2012). If diversity is unbounded, and either clade ages or diversification rates determine diversity in a region, then there should be a positive relationship between clade age and diversity (Rabosky, 2012). If, instead, diversity has an upper bound and is determined by diversity dependence and ecological limitations, then no relationship between age and diversity is expected (Rabosky, 2009; Cornell, 2013).
We conducted an extensive literature search and compiled data on species richness and age of colonization of the CFR (n =19) and the MB (n =20) for a total of 39 clades (Appendix S1). We excluded monotypic lineages. We performed separate analyses for both crown and stem ages, as not all publications reported both ages. To assess whether species richness increases with clade age we conducted Spearman rank correlations between clade age and log species richness. In addition, we compared the fit of three types of age–diversity models following the approach of Rabosky (2012): (1) a linear process-independent model, which is expected to fit the data well if species richness is determined by time since colonization alone and if extinction is low or if there is no pattern in the data; (2) a constant-rates birth–death model, which is expected to fit the data best if diversity is unbounded, and if time since colonization determines species richness in the presence of extinction; and (3) a logistic model, which is expected to fit the data best if diversification rates are diversity-dependent, meaning that diversity is asymptotic and determined by ecological limits. We fitted models and compared them with Akaike information criterion (AIC) scores using unpublished R code provided by Daniel Rabosky (University of Michigan, Ann Arbor). Analyses repeated for minimum, mean and maximum ages yielded congruent results, and we therefore report only the results for mean ages.
For the MB there was no relationship between clade age and species richness for either crown (n =20, Spearman's ρ = −0.057, P =0.81) or stem (n =14, ρ = 0.03, P =0.92) ages (Fig. 3, Table 3). The model with the lowest AIC score was the linear model, but this model was less than two AIC scores better than the logistic model, and therefore we are unable to favour one model over the other (Table 3). The slope of the linear model was not significantly different from zero (linear regression slope = −0.046, P =0.28), again revealing that there is no pattern in the MB data. The birth-death model (diversity is unbounded) was clearly rejected for the MB data as it had the highest AIC scores. For the CFR, there was a significant positive relationship between clade age and species richness for both crown (n =19, ρ = 0.75, P =0.0002) and stem (n =9, ρ = 0.63, P =0.067) ages. A constant-rates birth–death model was preferred over the linear and logistic models for the CFR crown age data (Table 3). For the stem age data, the birth-death model also had the lowest AIC score, but showed only a marginally better fit than the logistic model. However, given the fact that the n value of the stem age analysis is low, and that crown ages are more appropriate to track age-diversity dynamics (Rabosky et al. 2012), we favour the results of the crown age analysis. We interpret these results in the following sections.
Table 3. Results of age–diversity analyses. Spearman's rank correlation of plant clade age versus log number of species, and model fit statistics for the three age–diversity models (Rabosky, 2012) fitted to Cape Floristic Region (CFR) and Mediterranean Basin (MB) clades for both crown (CFR, n = 19; MB, n = 20) and stem ages (CFR, n = 9; MB, n = 14). For the CFR data, the model with the lowest AIC score is the birth–death model (no diversity limits), although this model is only marginally preferred to the logistic model for the CFR stem age data. For the MB data, the model with lowest AIC is the linear model, which is only marginally preferred to the logistic model for both crown and stem age data
Cape Floristic Region
Spearman's ρ (P-value)
Spearman's ρ (P-value)
Hypotheses Accounting for Disparities In Cape/ Mediterranean Diversity
Below, we examine the role of three main hypotheses that can be invoked to explain contrasting species richness patterns in the two regions (Fig. 4). We pay particular attention to groups that present a CFR–MB disjunction (Tables 1 & 2), as clades that conform to this unique biogeographical pattern remove the need to account for some confounding factors associated with phylogenetic constraints.
The clade age hypothesis
A potential explanation for the geographical differences in species richness between the two mediterranean-climate regions is that timings of colonization have varied between the two biomes (Fig. 4a). The CFR clades considered in our data set (Appendix S1) were significantly older than the MB clades (CFR crown mean = 18.3 Ma; MB crown mean = 7.09 Ma, P <0.001), suggesting that the higher diversity in the former region may have arisen because of the longer time available for clades to accumulate species. If a ‘time-for-speciation’ effect exists (Stephens & Wiens, 2003), there should be a positive correlation between clade age and species richness (Rabosky, 2012). For the combined data set (CFR + MB) considered in our analysis (Fig. 3) there was a strong positive relationship between clade age and richness for crown age (n =39, ρ = 0.49, P =0.002) and a moderate positive relationship for stem age (n =23, ρ = 0.31, P =0.145). This suggests that, overall, older clades tend to have more species, and provides support for the clade age hypothesis.
The time-for-speciation effect has rarely been examined for CFR–MB disjunct groups (Valente et al., 2011). Of the clades in Table 2, 6 out of 11 are older in the CFR, and, of these, Erica, Gladiolus, Lycium and Moraea are more diverse in this region, whereas Androcymbium–Colchicum and Hyacinthaceae are actually more diverse in the MB. This suggests that rates of diversification have been higher in the MB for the latter two groups. On the other hand, 5 of the 11 clades are older in the MB, and all of these are more diverse in this region, providing support for the clade age hypothesis, at least in these CFR–MB disjunct groups.
Importantly, the number of colonizations between regions in different periods may greatly influence diversity. Some CFR–MB disjunct groups, such as Carex (Escudero et al., 2012), Senecio sect. Senecio (Coleman et al., 2003) and Hyacinthaceae (Buerki et al., 2012), underwent one or more colonizations of the two regions, as inferred by phylogenetic information (Table 2), and thus may present several subclades with independent evolutionary histories.
The hypothesis of in situ diversification rates
The contrasting diversity patterns found within clades present in the two regions may also be due to non-homogeneous rates of diversification (speciation rate minus extinction rate) between the CFR and the MB (Fig. 4b). There are several a priori reasons to suggest that rates have varied (Kisel et al., 2011). On the one hand, the contrasting geomorphic and climatic histories of the CFR and the MB may have led to important differences in extinction rates (with rates potentially being higher in the MB). On the other hand, the differences in area, topography and ecological features may have caused heterogeneity of speciation rates.
Owing to the complexities associated with calculating rates of diversification (Ricklefs, 2007; Rabosky, 2009), few studies have reported such rates, and we therefore cannot provide a formal comparison of diversification rates between the two regions at this stage. However, the few studies that have estimated rates of diversification for CFR or MB clades provide a mixed picture. Some studies have found that rates of diversification in the CFR have been exceptionally high in some clades, such as Aizoaceae from the Greater Cape Floristic Region (Klak et al., 2004) and Phylica (Richardson et al., 2001). In contrast, other studies have showed that many CFR lineages have diversified at low to moderate rates (< 0.5 species Myr−1), such as Gladiolus (0.45 species Myr−1; Valente et al., 2011), Babiana (0.44 species Myr−1), Moraea (0.29 species Myr−1), Podalyrieae (0.15 species Myr−1) and Protea (0.22 species Myr−1; Schnitzler et al., 2011). In the MB, phylogenetically informed studies suggest that diversification rates have been very high (> 1 species Myr−1), rivalling or even exceeding the fastest plant radiations documented on the planet (Valente et al., 2010a). The most rapid MB radiations studied to date are Centaurea sect. Acrocentron (1.95 species Myr−1; Bell et al., 2012), Cistus (1.46–2.44 species Myr−1; Guzmán et al., 2009), Dianthus (2.2–7.6 species Myr−1; Valente et al., 2010a) and Tragopogon (0.84–2.71 species Myr−1; Bell et al., 2012).
Three studies have explicitly compared rates of diversification in taxa that occur in both the CFR and the MB. In Gladiolus, one of the largest clades in the CFR, overall rates of diversification were found to be moderate in that region (Valente et al., 2011). However, a particular clade of 28 species of Gladiolus endemic to the CFR was found to have diversified at 0.85–1.76 species Myr−1, three to four times more rapidly than the whole genus (0.24–0.45 species Myr−1) and three to five times more rapidly than the Gladiolus MB clade (0.16–0.31 species Myr−1; Valente et al., 2011). This difference was attributed, in part, to differences in the pollinator environment (Valente et al., 2012). The opposite scenario was found in Dianthus, a clade displaying an opposite diversity pattern to Gladiolus (i.e. more species in the MB). A large Eurasian lineage of Dianthus constituting the vast majority of MB taxa has diversified at the most rapid rate documented to date in plants (2.2–7.6 species Myr−1; Valente et al., 2010a), and at least 1.5 times more rapidly than the African lineage of Dianthus (1.4–5.4 species Myr−1; Valente, 2010). The reasons for the explosive diversification in Mediterranean Dianthus seem to be related to allopatric divergence driven by an unstable climate allied to unusually high rates of autopolyploidization and allopolyploidization (Weiss et al., 2002; Balao et al., 2010). The only other study that compared rates of diversification between the two hotspots was a phylogenetic analysis of Hyacinthaceae, a family with 184 species in the CFR and 262 species in the MB (Buerki et al., 2012). The study found high rates of speciation and extinction in both regions when compared with the rates in non-mediterranean biomes.
The hypothesis of spatio-ecological limits
A third hypothesis for explaining the differences in diversity between the CFR and the MB considers the role of ecological and spatial constraints to diversity. This hypothesis has received support from an increasing number of studies on a variety of taxonomic groups and regions that have found diversity to be regulated by diversity-dependent speciation–extinction dynamics (Phillimore & Price, 2008; Rabosky, 2009; Etienne et al., 2012; Rabosky et al., 2012). According to this idea, the relationship between clade age and species diversity is asymptotic (Fig. 4c): as a clade diversifies in a region, it progressively fills more niches, leading to a temporal decline in rates of speciation as fewer niches become available, until the speciation rate equals the extinction rate, at which point the clade has reached its diversity limit (Etienne & Haegeman, 2012; Cornell, 2013).
The role of ecological limits has not been explored explicitly in mediterranean plant radiations as it requires well-sampled phylogenies of clades in order to test for asymptotic behaviour in lineage-through-time (LTT) plots (Etienne et al., 2012). Well-sampled LTT plots of mediterranean clades are rare, and studies that did analyse LTT plots were unable to distinguish slowdowns or increases from sampling artefacts (Valente et al., 2010a; Schnitzler et al., 2011), or found a high number of patterns (Fiz-Palacios & Valcárcel, 2013).
An indirect way of investigating the existence of diversity limits is by testing the relationship between clade age and species diversity: if the diversity in a region depends significantly on the diversity carrying capacity/ecological limits for that area, then there should be no relationship between clade age and diversity (Rabosky, 2009, 2012). In our analysis of clade age versus species richness (Fig. 3, Table 3) we found a striking contrasting pattern between the CFR and the MB. There was no relationship between age and diversity for MB plant clades, which is consistent with previous studies that found a lack of relationship between clade age and species richness to be widespread in eukaryotic clades (Rabosky et al., 2012). The birth–death model (diversity is unbounded) was rejected. A lack of relationship between clade age and diversity cannot be explained by differential rates of diversification (as shown by simulations; Rabosky et al., 2012). Therefore, this suggests that speciation and extinction in the MB may be regulated by diversity-dependent processes, and that diversity may be bounded by ecological limits in this region. The fact that older clades in the MB do not exhibit more species than younger clades adds to the evidence that extinction may have been high in this region, as appears to be the case in northern temperate biomes (Tingley & Dubey, 2012).
In contrast, we found a significant positive relationship between clade age and richness in the CFR (Fig. 3, Table 3), and that a constant-rates birth–death model – expected to fit the data best if diversity is not determined by ecological limits – provides the best fit to the data. This suggests that diversity in this region may be unbounded and that, at least at this scale (< 50 Ma), many clades have not reached an asymptotic phase. In other words, older CFR clades continue to accumulate species and do not appear to have reached a diversity limit. This idea is further supported by the fact that 13 genera mostly from distantly related taxonomic families display over 100 species in the small area of the CFR (Goldblatt & Manning, 2002).
If diversity in the CFR is indeed unbounded, this greatly contrasts with the evidence we found for potential diversity limits in the MB. This leads to the question of what the mechanisms behind these differences may be. The existence and precise nature of ecological limits is a largely unresolved question, not only for mediterranean regions but also for other species-rich areas of the world (Cornell, 2013). Most arguments put forward to explain increases in diversity carrying capacity are related to increases in area (Kisel et al., 2011), and are not applicable in the case of a small hotspot such as the CFR. The evolution of key innovations that would allow species to escape competition and thus increase diversity carrying capacity has also been proposed (Etienne & Haegeman, 2012), but would not explain why such a large number of distantly related clades are so diverse in the CFR. We hypothesize that the CFR presents more potential for finer niche subdivision than the MB owing to its climatic stability (Cowling et al., 2004, 2005). High environmental reliability would allow small plant populations of species from diverse taxonomic backgrounds to adapt and persist for longer, even as niches became narrower and more partitioned. This would increase diversification rates by reducing rates of extinction, and would potentially raise the diversity carrying capacity, perhaps to a point at which diversity regulation becomes negligible. For this evolutionary mechanism to operate, ecological traits would need to be sufficiently labile to allow species to adapt quickly to these new finely divided niches (Cornell, 2013), a possibility that is supported by the fast rates of niche evolution documented in the CFR (Schnitzler et al., 2012). The fact that diversity in the CFR (small plants in small areas) is proportionally comparable only to that in the tropics (large plants and areas) suggests that these two biomes may support some of the highest ecological limits for plant species diversity.
The results of our age–diversity analysis must be viewed with caution, given that the sample sizes used are small, that molecular dating estimates are uncertain, and that the distribution ranges of the clades may have shifted dramatically over time. Further studies should aim to incorporate well-sampled phylogenies, full species distribution data and additional mediterranean floristic regions in order to test the effect of spatio-ecological limits to diversity in more detail.
We conclude that the processes proposed in all three hypotheses have played a role in explaining the observed higher density of species in the CFR than in the MB. On average, clades are older in the CFR and thus have had more time to accumulate species, which supports the clade age hypothesis. The hypothesis of higher diversification rates in the CFR appears to apply to some clades (e.g. Gladiolus), but clearly cannot account for diversity disparities in others (e.g. Dianthus). Finally, the diversity-limits hypothesis receives support from the contrasting age–diversity relationships found in the CFR (strong positive relationship) and the MB (no relationship), which suggests that diversity-dependent regulatory mechanisms in the CFR may be weak, whereas they appear to be of significant importance in the MB.
The emerging picture is that the remarkable species density in the CFR is attributable to the long-term persistence of lineages in a heterogeneous but stable ecological context. Environmental stability, low rates of extinction and the apparent lack of an upper bound to diversity in this biome have allowed the accumulation of thousands of plant species within an unusually small area. On the other hand, owing to its unstable climatic history, the MB has sustained high rates of extinction throughout most of its history, preventing the region from supporting and accumulating as many species as other more predictable and ancient environments. However, diversification rates in the MB have often exceeded those in the CFR (and in many cases those in other biodiversity hotspots) in the last few million years, and thus the MB may be acknowledged as a global hotspot for recent rapid speciation.
We thank Daniel Rabosky for providing code to test the fit of age–diversity models; M. Fernández-Mazuecos, A. Papadopulos, G. Mansion, A. Phillimore, O. Fiz, L. McInnes and J. Schnitzler for advice and/or data; and the editors and four referees for very constructive criticism of the manuscript. This research was funded by the European Commission Early Stage Network ‘Hotspots’ (FP6) and Marie-Curie Fellowhisp ‘Birdisland’ (FP7).
Luis M. Valente combines evolutionary theory with new biogeographical and genomic approaches to study the origin of species diversity in biodiversity hotspots and insular environments. He is currently using next-generation sequencing methods to study the causes of speciation in a rapid avian radiation in an archipelago.
Pablo Vargas leads projects on the evolution and molecular systematics of primarily Mediterranean plant groups. He has also been investigating the biogeography, colonization, phylogeography, speciation and microevolution of the floras of Macaronesia, Galápagos and Hawaii.