Pentaschistis (Poaceae) Diversity in the Cape Mediterranean Region

Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2009) 18, 586–595

Blackwell Publishing Ltd RESEARCH Pentaschistis (Poaceae) diversity in the PAPER Cape mediterranean region: habitat heterogeneity and climate stability C. Galley1*, H. P. Linder1 and N. E. Zimmermann2

1Institute of Systematic Botany, University of ABSTRACT Zurich, Zollikerstrasse 107, CH-8008 Zurich, Aim To evaluate the role of habitat heterogeneity on species richness and turnover Switzerland, 2Swiss Federal Research Institute in the mega species-rich Cape Floristic Region (Cape), the mediterranean region of WSL, Zürcherstrasse 111, CH-8903 Birmensdorf, Switzerland southern Africa. Location The Cape and Drakensberg regions of southern Africa. Methods Bioclimatic data were used to estimate habitat amount and habitat heterogeneity in the Cape and Drakensberg regions; these data were then used to explain the patterns of species diversity in the Pentaschistis clade (Poaceae) in these two regions. Habitat variables were used to create ‘bioclimatic units’ to characterize 1′ × 1′ cells in southern Africa and to describe the niches of species. Using these bioclimatic units, the niche and range sizes of species in the two regions were compared. A phylogram was used to test for age and lineage effects. Results Pentaschistis species diversity and turnover are higher in the Cape than the Drakensberg. There is no significant difference in the habitat amount and heterogeneity between the two regions. Species occupy the same range of estimated niche sizes, yet there are significantly more range-restricted Pentaschistis species in the Cape. Main conclusions The roles of age- and lineage-related effects are rejected; biodiversity differences lie in the regions. Current macrohabitat does not explain the differences in biodiversity between the two regions. The larger number of range-restricted species in the Cape cannot be explained by dispersal mechanism or the occupation of range-restricted habitats. Species of Pentaschistis and other Cape clades share characteristics associated with species from historically climatically stable areas, and palaeoclimatic and palaeontological evidence indicates the Cape climate has been more stable than the Drakensberg climate throughout the Pleistocene. We conclude that the corresponding lack of extinction might have allowed an accumulation of species in the Cape. Similar climatic and biological evidence for the south-west Australian Floristic and Mediterranean regions indicate that the same mechanism might explain the high species richness of these mediterranean regions. *Correspondence: C. Galley, Botany Keywords Department, Trinity College Dublin, Dublin 2, Ireland. Biodiversity, biogeography, Cape flora, extinction, macroecology, mediterranean, E-mail: [email protected] palaeoclimate, southern Africa.

(Cowling et al., 1996). Situated between c. 30°–40° north and INTRODUCTION south of the equator, these regions therefore present an exception Mediterranean regions harbour more plant species than their to the latitudinal gradient hypothesis (Rosenzweig, 1995). The neighbouring areas. They cover less than 5% of the Earth’s Cape Floristic Region (Goldblatt, 1978; hereafter referred to as surface yet contain almost 20% of the world’s plant species the ‘Cape’), the mediterranean region of southern Africa, is an

Habitats and climate stability in the Cape region extreme example of this with a species richness (9000 species in regions, nine subregions, which are well collected and for which 90,800 km2) comparable to mega species-rich tropical rainforests the species list is assumed to be complete, were chosen (Fig. 1). (Linder, 2003). Species richness and turnover between these subregions provide There have been several general explanations put forward to estimates of diversity at the landscape level. The subregions explain the high diversity in mediterranean regions, including within each region are dominated by a single vegetation type. those implicating habitat and topographic heterogeneity, winter rainfall and fire (Cowling et al., 1996). For the Cape, explana- Ecological datasets tions include those relating to the complex patterns of soil types combined with limited dispersal, the rainfall pattern, complex Raster datasets of climatic variables for southern Africa were fire regimes or the combined effects of limited dispersal and high obtained from the Schulze Atlas (Schulze, 1997). Variables that speciation rates (Goldblatt & Manning, 2002; Linder, 2003, and have been invoked to explain the high biodiversity in the Cape by references therein). An alternative set of hypotheses invokes past authors were chosen, plus those which from field observa- historical climates, including the relative climatic stability of tions were expected to be important in determining species the Cape during the Pliocene and/or Pleistocene, reducing distributions. The 13 variables used cover a range of factor types; extinction rates (Cowling & Lombard, 2002; Goldblatt & one relates to soils (soil fertility), one to topography (altitude), Manning, 2002). Cowling & Lombard (2002) demonstrated two to solar radiation and nine to climate (Table 1). The cell size that community diversity cannot explain the difference in species of all layers (1′ × 1′) corresponds to c. 1.6 × 1.4 km2. Data prepara- diversity within the Cape, and invoked different Pleistocene tion and spatial analyses were carried out in arcgis (ESRI, climatic histories across the region, affecting speciation and 1999–2002). The habitat variables that we considered are, by the extinction rates, to explain this. manner of the dataset, broad-scale and can be predicted at the There is no consensus regarding which habitat factors (Linder, landscape scale. 2003) and/or other factors (Savolainen & Forest, 2005) are the most important variables influencing biodiversity in the Cape. Specimen dataset An additional factor that must be considered is the history of the lineages of which a flora is comprised, including phylogenetic We obtained 2158 locality records for 63 of the 65 species of aspects and evolutionary time (Ricklefs & Schluter, 1993). Pentaschistis in the Cape and 359 locality records for the 11 species To target the question of species diversity in the Cape in this in the Drakensberg region from the Bolus Herbarium (University context, we use a clade which also occurs in a nearby area, the of Cape Town) and the National Herbarium (South African Drakensberg Region. Many clades occur in both the Cape and National Biodiversity Institute, Pretoria). We followed the Drakensberg regions and tend to be much more species poor in taxonomy used in Galley & Linder (2007). Subspecies were kept the latter (Hilliard & Burtt, 1987), although it is not generally distinct, but other infraspecific taxa were combined. Species that floristically impoverished (Carbutt & Edwards, 2006). The occur outside of the boundaries of the two regions, were omitted Pentaschistis clade [Pentameris P.Beauv., Pentaschistis Stapf, (Pentaschistis lima, Pentaschistis natalensis, Pentaschistis chippindal- Prionanthium Desv. and Pseudopentameris obtusifolia (Hochst.) liae). All localities were digitized to a 1′ × 1′ resolution, or used as Schweick.; hereafter referred to as ‘Pentaschistis’] comprises 65 given (from GPS recordings). The median number of records per species in the Cape and 11 species in the Drakensberg. It is a species was used to test for a collection bias between the Cape common element in both regions and is the most species-rich and Drakensberg. grass lineage in the Cape flora. A phylogeny of Pentaschistis was recently published and the taxonomy is relatively well Characterising the regions and sub-regions known (Galley & Linder, 2007), allowing us to use comparable taxonomic units. Habitat gradients within the Cape and Drakensberg regions were This study aims to examine the role of habitat heterogeneity compared for the 13 variables (Table 1). The lengths of the in the species richness and turnover of the Cape, using the habitat gradients were estimated from the standard deviations Drakensberg as a comparative area; a rate-corrected phylogram for each of the variables, and these were compared between the is used to test for lineage- and age-related effects. two regions. The local steepness of habitat gradients was estimated with neighbourhood statistics for the same 13 variables as follows. Using neighbourhoods of 3 × 3 cells, the range of METHODS values in the neighbourhood was calculated and assigned to each central cell. The means of the ‘neighbourhood range’ The regions distributions for the Cape and the Drakensberg were compared The study areas are in southern Africa: at the south-western tip, for each variable. the Cape, and in the east, the Drakensberg region (Fig. 1) To estimate the amount of multivariate habitat in each sub- (Rutherford & Westfall, 1986). The Cape is dominated by two region, a single ordination based on 10 variables (Table 1; highly main vegetation types, fynbos and renosterveld, with the majority correlated and categorical variables removed) was calculated in of Pentaschistis species occurring in the former. The Drakensberg ntsyspc (Rohlf, 2000–2003) (results not shown). The ‘volume’ region is dominated by grassland. Within each of the two of the cloud of points of each subregion was estimated using the

C. Galley et al.

Figure 1 Map of the study area showing the regions outlined by black lines (the Cape to the left, the Drakensberg to the right) and subregions numbered and shown in white: 1, Cederberg; 2, Hexrivierberg; 3, Langeberg; 4, Cape Peninsula; 5, Mount aux Sources; 6, Cathedral Peak; 7, Cathkin Peak; 8, Sani Pass to Bushman’s Nek; 9, Naudes Nek to Lundean Pass. Specimen localities are shown by black points and the background depicts altitude (white to darkest grey: 0 to 3800 m). The Cape boundary was taken from extent of the broad habitat units (Cowling and Heijnis, 2001). The extent of the Drakensberg region is the total extent of following vegetation types from the vegmap beta 4.0 (National Botanical Institute, South Africa): Southern Drakensberg Highland grassland, Northern Drakensberg Highland grassland, Lesotho Highland Basalt grassland, Drakensberg-Amatole Afromontane fynbos, Drakensberg Afroalpine heathland, Drakensberg montane shrubland, Drakensberg wetlands, Stormberg Plateau grassland, Ukhalhlamba basalt grassland, Western Lesotho basalt shrubland, Drakensberg wetlands and Lesotho mires. equation (4/3π)xyz, where x, y and z are the radii derived from variables (Appendix S1 in Supporting Information) and a the extent of each cloud of points along the first three axes of the discriminant function analysis (DFA) was used to indicate which ordination. The central 90% of data were used to minimize variables are important in separating the species within the two outlier effects. To adjust for subregion size, the volume was regions (Appendix S1). Five variables were chosen which both divided by the number of 1′ × 1′ points (approximately the area). contributed substantially to the first four axes in the DFA and To contrast the species richness between the subregions of which are least correlated with each other. These were: potential the Cape and the Drakensberg, the standardized residuals from evaporation, continentality, degree days in the summer, frost a regression of log10 species to log10 area for each subregion days and mean annual precipitation. The five variables were were compared. Species turnover between subregions was transformed to normalize their distribution and were cate- calculated using the Bsim value, which measures species turnover gorized as to maximize the discriminatory power of the variable independent of any species richness gradient. Bsim is calculated (Table 1). These categorized variables were concatenated to make as: min (b, c) × [min(b, c) + a]–1, where a is the total number of a five-digit code, the BU. The inclusion of bedrock, which has species present in both areas, b is the number of species present been implicated in several explanations of species diversity in the in the second area but not the first and c is the number of Cape, was tested separately (its suitability could not be tested species present in the first area but not in the second (Lennon using the DFA as it is a non-ordered variable); it was initially et al., 2001; Koleff et al., 2003). included in the BUs, replacing the poorest performing variable, degree-days in the summer. These BUs performed worse in predicting species distributions than a purely bioclimatic dataset, Characterizing the species thus the latter was used for analysis. Species were characterized by combining categorical environ- Each 1′ × 1′ cell in southern Africa was assigned a BU. As the mental variables to create ‘bioclimatic units’ (BUs). Dataset size plant records were also assigned to these cells, the BU for each constraints meant only five variables could be used for these BUs. specimen could be determined, and from this the bioclimatic A correlation matrix was used to identify highly correlated envelope of each species was inferred.

Habitats and climate stability in the Cape region

Table 1 Thirteen variables were used to estimate habitat gradient length and local habitat steepness. The ten variables used in the ordination (to estimate the habitat volume of the subregions) and the ﬁve used to construct the bioclimatic units (BUs; see text) are indicated. APAN, mean annual patential evaporation, class A-pan equivalent model; MAP, mean annual precipitation.

Brief description (refer to Schulze, 1997, Used in the Used to Transformation for extra details) ordination? construct the BUs? for BUs Categorization for BUs

Altitude Elevation above sea level No No – – APAN Potential evaporation derived from Yes Yes none Nine equal intervals monthly means of: daily max. air temperature, solar radiation, precipitation, also taking into account the month, altitude and the evaporation region (of 12 in southern Africa) Continentality Derived from: January max. temperature Yes Yes None Eight equal intervals, except minus July min. temperature highest interval is larger (contains very few records) Coefﬁcient of Interannual variation in rainfall; utilizing Yes No – – variation of annual the relationship of MAP and known precipitation (CVAP) annual variability, CVAP is determined by an equation involving MAP and parameter values which vary with climatic region (summer, winter or all year rainfall) Degree days (DD) Cumulative number of degree days Yes No – – winter between April and September, above a threshold of 10 °C Degree days (DD) As above, between October and March No Yes Log(10) Seven categories based on summer natural breaks (Jenk’s algorithm) Frost days Number of days per year with heavy frost Yes Yes None Nine categories, log(2) Mean annual Using data from 6000 rainfall stations and Yes Yes Square root Eight equal intervals, except precipitation integrating over 34 regions homogeneous the highest interval is larger for rainfall ‘controlling’ factors (contains very few records) Rain concentration Ranges from 0 (precipitation is equal in Yes No – – all months) to 100% (all precipitation falls in 1 month); estimated by Markham’s method (see Schulze, 1997) Rain seasonality Categorical description of when the most Yes No – – rain falls (ﬁve categories) or all year rainfall Soil fertility Product of clay content and base status No No – – Solar radiation Estimated from extraterrestrial solar Yes No – – (January) radiation, air temperature and daily temperature range of an area (using Clemence’s method; see Schulze, 1997), for January averages Solar radiation (July) As above, for July averages Yes No – –

The geographical range of each species was estimated using the The average of these numbers was used to obtain a single value extent of occurrence (sensu Gaston, 1994), and the distributions representing the ‘occurrence of habitat’ for each species. The of species ranges in the Drakensberg and Cape were compared. occurrence of habitat and the geographical range of Cape species were compared. Correlates of range-restricted species Disentangling clade history and ecological relatedness BUs were applied to each 1′ × 1′ cell in southern Africa. To test whether range-restricted species occupy rare habitats, the BUs The DNA matrix from Galley & Linder (2007), with the addition assigned to each Cape species were identiﬁed and the number of of new sequences from Pentaschistis praecox and Pseudopentameris cells in the Cape that have each of these BUs was determined. obtusifolia, was used to obtain a rate-corrected phylogeny (analyses

C. Galley et al.

Figure 2 The nine subregions from the Drakensberg (left) and the Cape (right) regions, showing Pentaschistis species richness in each subregion in brackets (in bold). The

numbers between the subregions are the Bsim values (see text) on a scale representing no or low species turnover (0) to complete species turnover (1) between any two subregions. as described in Galley & Linder, 2007). This phylogram plus optimized species richness among the subregions (d.f. = 8, r2 = 0.462, ancestral areas (Galley et al., 2007) were used to estimate the P = 0.044); the standardized residuals from this regression show branch lengths of lineages (including internal branches) in the that the subregions from the Cape tend to be more species rich

Drakensberg and the Cape. Non-focal area taxa were removed, (t-test, d.f. = 7, P = 0.016). Bsim values between subregions in the and the two sets of branch lengths were compared to test whether Cape are much higher than in the Drakensberg (Fig. 2), demon- the lineages in the Drakensberg tend to be younger. strating that species turnover is much higher in the former (there A Mantel test was used to test if more closely related species are four species which are common to all the subregions in the tend to occupy similar niches. For each habitat variable (Table 1), Drakensberg thus the Bsim statistic is ‘0’ across the subregions). data from individual collections were used to construct 0.1 and Overall, the subregions from the two regions have similar 0.9 quantiles for each species (quantiles were used to reduce the habitat volume (size adjusted, t-test on log-transformed data, effect of outliers). From each of these two matrices, following d.f. = 7, P = 0.596). Correspondingly, the volume of habitat in standardization, a dissimilarity matrix was calculated (average the subregions does not explain the variation in species richness taxonomic distance as implemented in ntsyspc) to estimate the (Fig. 3, r2 = 0.244, P = 0.160). average ecological distance between all species pairs, across all variables. A pairwise matrix of patristic distances (i.e. genetic Characterizing the species distance taking phylogeny into account) was calculated to estimate phylogenetic relatedness (paup; Swofford 2002). The The number of BUs occupied by each species does not differ Mantel test was carried out between the phylogenetic matrix between the two regions (t-test on log-transformed data, and each of the two ecological matrices, using 250 permutations d.f. = 69, P = 0.627). The geographical ranges of species in the (ntsyspc). Cape tend to be smaller than those of the Drakensberg (MWU-test, All statistical tests were carried out in spss (SPSS Inc., 2003). d.f. = 69, P = 0.036, see Fig. 4). In the Cape, 13 of the 65 species Data between the two regions were compared using a regression have ranges of < 100 km2; three of these have been collected from or the Student’s t-test, where distributions were normal or where multiple localities. By contrast, the smallest geographical range data could be transformed to achieve this. Otherwise, we applied of a Drakensberg species is c. 500 km2. Additionally, the distribu- a Mann–Whitney U-test (‘MWU-test’ hereafter). In all tests, we tion of the range sizes of Cape species is significantly skewed concluded significance based on α = 0.05. Species with outlying (skewness statistic = 3.783, SE = 0.314, single-collection species variables were omitted from individual tests. removed) whereas the Drakensberg distribution is not (skewness statistic = 0.906, SE = 0.687). The geographical ranges of Cape species and their occurrence RESULTS of habitat are not correlated (r2 = 0.03, P = 0.256). The specimen dataset Clade history and ecological relatedness The median number of records for the two regions do not differ significantly (Cape: median of 17.5 specimens per species (s/s); Five to ten migrations between the Cape and Drakensberg are Drakensberg: 30.5 s/s; d.f. = 74, MWU-test, P = 0.293). required to explain the distribution of Pentaschistis (see also Galley et al., 2007). The lineages in the Drakensberg do not represent especially recent migrations as the first lineage to occupy the Characterizing the regions and subregions Drakensberg is sister to a larger Cape clade (in clade I, Fig. 5) and On average, habitat gradients and the habitat steepness are the second occupation of the Drakensberg occurred shortly after similar in the Cape and the Drakensberg. The Cape has longer the diversification of the main clade in the Cape (Fig. 5; see also habitat gradients for seven variables, the Drakensberg for six; Galley et al., 2007). Additionally, the lineages in the Drakensberg the Cape has steeper habitat gradients for seven variables, the and the Cape do not differ in branch lengths (t-test on log- Drakensberg for six (Appendix S2). transformed data, d.f. = 134, P = 0.104). The subregions of the Cape contain more species than the There is no correlation between the ecological distance of Drakensberg subregions. A linear regression of log species = log species and their patristic/phylogenetic distance (lowest 10% area explains a significant but small amount of the variation in quantiles, r2 = 0.035; uppermost 10% quantiles, r2 = 0.047).

Figure 4 Histogram of species ranges with Cape species in white and Drakensberg species in black.

would result in an increase in the apparent species ranges and consequently a decrease in the already very low species turnover. Although species diversity would increase if new species were discovered, an unrealistically large number of discoveries would Figure 3 Scatterplot showing the log species richness of the nine be necessary to redress the balance in species diversity between subregions against their estimated habitat ‘volume’ adjusted by size. the two regions. Additionally, to mitigate the effects of any bias Here, habitat volume is the volume of the cloud of points of each between regions, we concentrated many analyses on nine subregion as estimated using the equation (4/3π)xyz, where x, y and well-collected subregions which should minimize false absences. z are the radii derived from the extent of each cloud along the first three axes of an ordination on 10 climate variables (see Table 1). The central 90% of data were used to minimize outlier effects. To adjust Explaining the difference in diversity ′ × ′ for subregion size, the volume was divided by the number of 1 1 There are more Pentaschistis species in the Cape than in the points (approximating to area). Drakensberg, at both regional and sub-regional levels, correcting for area. There is a higher proportion of range-restricted Pentaschistis species in the Cape, reflected by the higher species turnover DISCUSSION between the subregions in this region, compared with in the Drakensberg. Differences in the species richness between regions Some methodological issues could be a consequence of the time available for diversification, if It is important to quantify the comparative ecology of species in the Drakensberg had been occupied more recently than the mega-diverse regions and GIS is a suitable and cost-effective tool Cape. Using the phylogram, we reject this explanation: the to characterize habitats. However, the accuracy of habitat charac- diversification rate of Drakensberg lineages is lower. A second terization and predictive modelling is limited by the number of explanation might be rapid diversification following a key occurrence points per species (Stockwell & Peterson, 2002). Our innovation in the Cape, as had been argued for the Rushioideae dataset contains a mix of species represented by many specimens (Klak et al., 2004). This would, however, require loss of a key and by a few specimens; of the latter, some are very rare, whereas innovation with each of the five to ten dispersals into the others are common or widespread but poorly collected. Even if Drakensberg, an unparsimonious solution. An explanation poor species collection is the problem, given the time and cost of related to the region and not to the taxa is required. Such an collecting biological data this will not be remedied in the near external factor(s) seems more likely as several Cape-based lineages future. Discarding species with few records to reduce error with species in other areas (including the Drakensberg) are much associated with small samples is one solution, but here would more species rich in the Cape (Hilliard & Burtt, 1987). miss out what are potentially the most important species, i.e. the range-restricted species. We used coarsely categorized variables Present environment: habitat heterogeneity to construct the BUs, which decreases their precision but which has been shown to increase the success of predictive modelling Various habitat parameters have been central to hypotheses when dealing with small datasets (Stockwell & Peterson, 2002). regarding the causes of the high species richness in the Cape Relative collecting effort between the two regions might affect (Linder, 1991; Cowling et al., 1996; Goldblatt & Manning, 2002), the conclusions reached. The Cape flora, is, compared to the in southern Africa (Thuiller et al., 2006) and globally (Kreft & Drakensberg, a generally well-collected flora (Linder, 1996). Jetz, 2007). However, at neither the regional nor subregional However, potential undersampling in the Drakensberg would scale is there a significant difference in habitat richness between not affect the conclusions reached here; increased collections the Cape and the Drakensberg, nor a consistent difference in the

Figure 5 Ultrametric tree of the Pentaschistis clade indicating species found in the Cape, Drakensberg or other regions. local steepness of habitat gradients. Consequently, our environ- range-restricted species in general compared with other areas of mental data do not explain the higher species richness or species sub-Saharan Africa (Carbutt & Edwards, 2006), and the local turnover of the Cape. This is consistent with Cowling & Lombard richness of these species correlates with the local species diversity (2002), who used plant community heterogeneity as a surrogate (Cowling & Lombard, 2002) and species turnover (Cowling, for environmental heterogeneity and found this insufficient 1990) in general across the Cape. Consequently, they might be the to explain differences in species diversity within the Cape. The key to understanding the high species richness in the Cape flora. predictive model of Thuiller et al. (2006) of species richness in Species might be range restricted because they are young southern Africa used topographic heterogeneity as the most (Willis, 1922), have limited seed dispersal, occur in rare habitats important predictor of species richness; notably, it over-predicts or have reduced extinction rates (Gaston & Blackburn, 2000; the species richness in the Drakensberg, predicting it to be Woodward & Kelly, 2003). There is no apparent difference in the comparable to that of the Cape. diaspores of range-restricted and non-range-restricted species of Pentaschistis, and we demonstrate that range-restricted species do not tend to occupy rare habitats. The age of species is impossible Species niche sizes and geographical ranges to estimate from species-level sampling and therefore either this, Given that the habitat volumes of the two regions are similar, or a reduced rate of extinction, allows more range-restricted the higher species richness and turnover in the Cape might be species to survive in the Cape. a consequence of species having either smaller niches or being more range restricted. Using BUs as a proxy for niche size, we Historical environment: climate stability and extinction reject the first of these explanations. There are more range- restricted Pentaschistis species in the Cape, explaining at least Range-restricted species are particularly sensitive to climate change- part of the higher species richness and turnover of this region. driven extinction (Dynesius & Jansson, 2000) and correspondingly, The Cape is also known to have an over-representation of a relatively stable climatic history has been used to explain several

592 Global Ecology and Biogeography, 18, 586–595, © 2009 Blackwell Publishing Ltd Habitats and climate stability in the Cape region diversity hotspots (e.g. African forests, Fjeldså & Lovett, 1997; Pentaschistis capensis growing in running water; Pentaschistis sub-Saharan African flora, Linder, 2001). Climatic stability may densifolia on moss beds on rock shelves) whereas specialists are be as or more important than the present environment in rare in the Drakensberg. Importantly, these characteristics are explaining levels of endemism (Jansson, 2003). Relative climatic found in Cape plant taxa in general; species turnover and the stability has been proposed as an explanation for the high species proportion of range-restricted species is very high in the Cape, richness of the Cape (Goldblatt & Manning, 2002) and has been and there is an abundance of species that occupy very specialized used to explain the discrepancy between species diversity in the niches (Goldblatt & Manning, 2000) or are poor dispersers western and eastern Cape (Cowling & Lombard, 2002). (Goldblatt, 1997). We conclude that a lack of extinction in glacial There is evidence of greater climatic fluctuation, during the periods allowed an accumulation of species in the Cape region Pleistocene and particularly the Last Glacial Maximum (LGM), leading, in part, to its higher species diversity. in the Drakensberg compared with the Cape. Around the Cape region, LGM maximal cooling was 5–6 °C (Congo Caves, Other Mediterranean areas Outshoorn, Vogel, 1983; Uitenhage Aquifer, Heaton et al., 1986). In the Drakensberg, the situation is less clear, but there is periglacial The Cape’s flora shares high species richness, endemism and evidence (Tyson, 1986) indicating drops of between 5.5 °C and rarity with floras of other mediterranean regions (Greuter, 1991; 8 °C during the late Pleistocene (Harper, 1969). This difference is Cowling et al., 1996). Such floras may have been important for supported by global circulation models (GCMs) that show plant survival during glacial/inter-glacial cycling, being situated comparatively less cooling in the western Cape than in the between higher latitudes, which became glaciated, and deserts, Drakensberg in the LGM (Weaver et al., 1998). which expanded with increased aridity (Hewitt, 2004). There is In the Cape, palaeontological data and climatic modelling evidence of climatic stability in several of these regions and we indicate that fynbos, presumably with the associated Pentaschistis suggest that lowered extinction might have also played a role in species, maintained or expanded its range during the LGM assembling these species-rich floras. (Partridge et al., 1999). Ericaceae and Restionaceae pollen on the The South Western Australia Floristic Region lies near the Walvis Bay Ridge indicates presence of the winter rainfall regime coast, outside the limits of the estimated arid zone of the LGM north of its current location during the LGM (Shi et al., 2000). (Crisp et al., 2001). The similarity of its Pliocene and current Pollen data from middens over the last 19,700 years at the flora suggests relatively moderate impacts of climate change on north-east of the Cederberg Mountains show that fynbos lineage persistence (Hopper & Gioia, 2004), and the high endemism vegetation prevailed at the site for the whole period (Scott, in this region contrasts with that of the central Australian flora, 1994). A northward shift of fynbos vegetation has also been attributed to an unstable climate in this region (Crisp et al., demonstrated by modelling vegetation distributions (Midgley 2001). The Mediterranean region contains almost a tenth of the et al., 2001; this also shows a reduction in fynbos in southerly world’s vascular plants (Greuter, 1991) and 50% endemism areas but importantly with persisting areas of fynbos). In the (Cowling et al., 1996). Like the Cape, the high species richness in Drakensberg, palaeoenvironmental data were used to estimate this region is due to many range-restricted species rather than to that steppe vegetation occupied significant areas around the high alpha-diversity (Greuter, 1991). Despite climatic instability Lesotho highlands and mountains of East Africa during the LGM during glacial periods, numerous refugia that allowed long-term (Partridge et al., 1999). Lower down, the extent of Afromontane species persistence have been identified; these also provided forest estimated by GCMs to have remained through the LGM, is source material for the reoccupation of previously glaciated areas very small (Eeley et al., 1999), indicative of considerable climatic (Taberlet et al., 1998). Palaeoecological data suggest that central change in the region. coastal California experienced climatic change during the Aside from depleting biodiversity, climate change may also Pleistocene but that a mediterranean climate persisted in the region influence the attributes of extant species (Dynesius & Jansson, throughout the Quaternary, thus it could have been an ‘ice-age 2000). Regions that have been relatively climatically stable tend refugium’ for cold-sensitive plants (Johnson, 1977). Evidence to have higher species richness, higher species turnover, more for climatic stability in the Chilean mediterranean region is less range-restricted species, more habitat specialists and more clear, but notably the matorral flora is the least species-rich of the species that are poor dispersers (Dynesius & Jansson, 2000; five regions and has the fewest endemics (Cowling et al., 1996). Lawes et al., 2007). These specialist, small-ranging and poorly dispersing species are those predicted to go extinct in the face of ACKNOWLEDGEMENTS climate change. Such ‘climatic filtering’ has been demonstrated for Afrotemperate forest fauna in the east of South Africa, in con- C.G. and H.P.L. would like to thank the University of Zurich, the trast to the fauna of the scarp forests which were less decimated Swiss National Science Foundation (3100–066594), the Georges (Lawes et al., 2007). For Pentaschistis, a comparison between the und Antoine Claraz-Schenkung and the Schweizerische Akademie Drakensberg and Cape fits these expectations well: species der Naturwissenschaften Travel Grant for funding, and Cape richness and turnover are higher in the Cape than the Drakensberg, Nature for collecting permission. N.Z. would like to thank the and there is a higher proportion of range-restricted species in the FP6 program Ecochange (GOCE-CT-2007-036866) for funding. Cape. Many Pentaschistis species in the Cape are habitat specialists We would also like to thank three anonymous referees for their (e.g. Pentaschistis calcicola on exposed limestone platforms; constructive comments.

Goldblatt, P. & Manning, J.C. (2002) Plant diversity of the Cape REFERENCES region of Southern Africa. Annals of the Missouri Botanical Carbutt, C. & Edwards, T.J. (2006) The endemic and near-endemic Garden, 89, 281–302. angiosperms of the Drakensberg Alpine Centre. South African Greuter, W. (1991) Botanical diversity, endemism, rarity, and Journal of Botany, 72, 105–132. extinction in the Mediterranean area: an analysis based on the Cowling, R.M. (1990) Diversity components in a species-rich published volumes of Med-Checklist. Botanika Chronika, 10, area of the Cape Floristic Region. Journal of Vegetation Science, 63–79. 1, 699–710. Harper, G. (1969) Periglacial evidence in southern Africa during Cowling, R.M. & Heijnis, R.H.M. (2001) The identification of the Pleistocene epoch. Palaeoecology of Africa, 4, 71–101. Broad Habitat Units as biodiversity entities for systematic con- Heaton, T.H.E., Talma, A.S. & Vogel, J.C. (1986) Dissolved gas servation planning in the Cape Floristic Region. South African paleotemperatures and 18O variations derived from groundwater Journal of Botany, 67, 15–38. near Uitenhage, South Africa. Quaternary Research, 25, 79– Cowling, R.M. & Lombard, A.T. (2002) Heterogeneity, speciation/ 88. extinction history and climate: explaining regional plant Hewitt, G.M. (2004) Genetic consequences of climatic oscilla- diversity patterns in the Cape Floristic Region. Diversity and tions in the Quaternary. Philosophical Transactions of the Royal Distributions, 8, 163–179. Society B: Biological Sciences, 359, 183–195. Cowling, R.M., Rundel, P.W., Lamont, B.B., Arroyo, M.K. & Hilliard, O.M. & Burtt, B.M. (1987) The botany of the southern Arianoutsou, M. (1996) Plant diversity in Mediterranean- Natal Drakensberg. NBI, Pretoria. climate regions. Trends in Ecology and Evolution, 11, 362–366. Hopper, S.D. & Gioia, P. (2004) The Southwest Australian Floristic Crisp, M.D., Laffan, S., Linder, H.P. & Monro, A. (2001) Region: evolution and conservation of a global hot spot of Endemism in the Australian flora. Journal of Biogeography, 28, biodiversity. Annual Review of Ecology Evolution and Systematics, 183–198. 35, 623–650. Dynesius, M. & Jansson, R. (2000) Evolutionary consequences Jansson, R. (2003) Global patterns in endemism explained by of changes in species’ geographical distributions driven by past climatic change. Proceedings of the Royal Society B: Biological Milankovitch climate oscillations. Proceedings of the National Sciences, 270, 583–590. Academy of Sciences USA, 97, 9115–9120. Johnson, D.L. (1977) The late Quaternary climate of coastal Eeley, H.A.C., Lawes, M.J. & Piper, S.E. (1999) The influence of California: evidence for an ice age refugium. Quaternary climate change on the distribution of indigenous forest in Research, 8, 154–179. KwaZulu-Natal, South Africa. Journal of Biogeography, 26, Klak, C., Reeves, G. & Hedderson, T. (2004) Unmatched tempo 595–617. of evolution in Southern African semi-desert ice plants. ESRI Inc. (1999–2002) ArcGIS, Desktop version 8.3.0. ESRI, Nature, 427, 63–65. Redlands, CA. Koleff, P., Gaston, K.S. & Lennon, J.L. (2003) Measuring beta Fjeldså, J. & Lovett, J.C. (1997) Geographical patterns of old and diversity for presence–absence data. Journal of Animal Ecology, young species in African forest biota: the significance of 72, 367–382. specific montane areas as evolutionary centres. Biodiversity Kreft, H. & Jetz, W. (2007) Global patterns and determinants of and Conservation, 6, 325–346. vascular plant diversity. Proceedings of the National Academy of Galley, C. & Linder, H.P. (2007) The phylogeny of the Pentaschistis Sciences USA, 104, 5925–5930. clade (Danthonioideae, Poaceae) based on chloroplast DNA, Lawes, M.J., Eeley, H.A.C., Findlay, N.J. & Forbes, D. (2007) and the evolution and loss of complex characters. Evolution, Resilient forest faunal communities in South Africa: a legacy 61, 864–884. of palaeoclimatic change and extinction filtering? Journal of Galley, C., Bytebier, B., Bellstedt, D.U. & Linder, H.P. (2007) The Biogeography, 34, 1246–1264. Cape element in the Afrotemperate flora: from Cape to Cairo? Lennon, J.J., Koleff, O., Greenwood, J.J.D. & Gaston, K.J. (2001) Proceedings of the Royal Society B: Biological Sciences, 274, 535– The geographical structure of British bird distributions: 543. diversity, spatial turnover and scale. Journal of Animal Ecology, Gaston, K.J. (1994) Measuring geographic range sizes. Ecography, 70, 966–979. 17, 198–205. Linder, H.P. (1991) Environmental correlates of patterns of Gaston, K.J. & Blackburn, T.M. (2000) Pattern and processes in species richness in the south-western Cape Province of South macroecology. Blackwell Scientific, Oxford. Africa. Journal of Biogeography, 18, 509–518. Goldblatt, P. (1978) An analysis of the flora of Southern Africa: Linder, H.P. (1996) Priorities for taxonomic research on the its characteristics, relationships, and origins. Annals of the Cape flora. South African Journal of Science, 92, 323–326. Missouri Botanical Garden, 65, 369–436. Linder, H.P. (2001) Plant diversity and endemism in sub- Goldblatt, P. (1997) Floristic diversity in the Cape Flora of South Saharan tropical Africa. Journal of Biogeography, 28, 169–182. Africa. Biodiversity and Conservation, 6, 359–377. Linder, H.P. (2003) The radiation of the Cape flora, southern Goldblatt, P. & Manning, J.C. (2000) Cape plants. A conspectus of Africa. Biological Reviews, 78, 597–638. the Cape flora of South Africa. National Botanical Institute, Midgley, G.F., Hannah, L., Roberts, R., MacDonald, D.J. & Pretoria. Allsopp, J. (2001) Have Pleistocene climatic cycles influenced

species richness patterns in the greater Cape Mediterranean Weaver, A.J., Eby, M., Fanning, A.F. & Wiebe, E.C. (1998) Region? Journal of Mediterranean Ecology, 2, 137–144. Simulated influence of carbon dioxide, orbital forcing and ice Partridge, T.C., Scott, L. & Hamilton, J.E. (1999) Synthetic sheets on the climate of the Last Glacial Maximum. Nature, reconstructions of southern African environments during the 394, 847–853. Last Glacial Maximum (21–18 kyr) and the Holocene Altithermal Willis, J.C. (1922) Age and area: a study in geographical dis- (8–6 kyr). Quaternary International, 57/58, 207–214. tribution and origin of species. Cambridge University Press, Ricklefs, R.E. & Schluter, D. (1993) Species diversity: regional Cambridge. and historical influences. Species diversity in ecological com- Woodward, F.I. & Kelly, C.K. (2003) Why are species not munities: historical and geographical perspectives (ed. by R.E. more widely distributed? Physiological and environmental Ricklefs and D. Schluter), pp. 350–363. Chicago University limits. Macroecology: concepts and consequences (ed. by Press, Chicago. T.M. Blackburn and K.J. Gaston), pp. 239–255. Blackwell Rohlf, F.J. (2000–2003) NTSYSpc: numerical taxonomy and Publishing, Oxford. multivariate analysis system, Version 2.11S, Exeter Publishing, New York. SUPPORTING INFORMATION Rosenzweig, M.L. (1995) Species diversity in space and time. Cambridge University Press, Cambridge. Additional Supporting Information may be found in the online Rutherford, M.C. & Westfall, R.H. (1986) Biomes of Southern version of this article: Africa – an objective categorisation. Memoirs of the Botanical Survey of South Africa, 54, 1–98. Appendix S1 Correlation matrices and standardized canonical Savolainen, V. & Forest, F. (2005) Species-level phylogenetics discriminant function coefficients of environmental variables. from continental biodiversity hotspots. Plant species-level systematics: new perspectives on pattern and process (ed. by Appendix S2 Environmental variables and statistics indicating F.T. Bakker, L.W. Chatrou, B. Gravendeel and P.B. Pelser), the length of habitat gradients. pp. 17–30. A.R.G. Gantner Verlag, Ruggell, Liechtenstein. Schulze, R.E. (1997) South African atlas of agrohydrology and Please note: Wiley-Blackwell is not responsible for the content climatology. Report TT82/96. Water Research Commission, or functionality of any supporting materials supplied by the Pretoria. authors. Any queries (other than missing material) should be Scott, L. (1994) Palynology of late Pleistocene hyrax middens, directed to the corresponding author for the article. southwestern Cape Province, South Africa: a preliminary report. Historical Biology, 9, 71–81. BIOSKETCHES Shi, N., Dupont, L.M., Beug, H.-J. & Schneider, R. (2000) Cor- relation between vegetation in southwestern Africa and oceanic Chloé Galley’s research interests lie in understanding upwelling in the past 21,000 years. Quaternary Research, 54, patterns of species diversity using a combination of 72–80. phylogenetics, historical biogeography and macroecology. SPSS Inc. (2003) SPSS, Version 12.0.1. SPSS, Chicago. She is also interested in how landscape ecology affects Stockwell, D.R.B. & Peterson, A.T. (2002) Effects of sample size species composition. on accuracy of species distribution models. Ecological Modelling, Peter Linder is interested in the patterns and processes 148, 1–13. of taxonomic and geographical diversification of the Swofford, D.L. (2002) PAUP* (phylogenetic analysis using parsi- austral floras, with a focus on the Cape flora of southern mony and other methods), Version 4.10b. Sinauer Associates, Africa. He is also investigating the diversification and Sunderland, MA. systematics of Restionaceae and the grass subfamily Taberlet, P., Fumagalli, L., Wust-Saucy, A.G. & Cosson, J.F. Danthonioideae, in time and space. (1998) Comparative phylogeography and postglacial colo- Nick Zimmermann’s research interests lie in vegetation nization routes in Europe. Molecular Ecology, 7, 453–464. ecology, vegetation dynamics and plant species diversity at Thuiller, W., Midgley, G.F., Rouget, M. & Cowling, R.M. (2006) various spatial and temporal scales. His research has Predicting patterns of plant species richness in megadiverse recently focused on modelling biophysical constraints of South Africa. Ecography, 29, 733–744. species and community distribution, and on species Tyson, P.D. (1986) Climatic change and variability in southern competitive dynamics. Africa. Oxford University Press, Cape Town. Vogel, J.C. (1983) Isotopic evidence for the past climates and vegetation of southern Africa. Bothalia, 14, 391–394. Editor: Jeremy Kerr