Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Heterogeneity, speciation/extinction history and climate: explaining regional plant diversity patterns in the Cape Floristic Region

Heterogeneity, speciation/extinction history and climate: explaining regional plant diversity... Introduction Variation in plant diversity at the regional scale (0.1–10 6 km 2 ) within and between biogeographical zones is explained most frequently by models that incorporate as explanatory variables measures of climate/energy ( Richerson & Lum, 1980 ; Currie & Paquin, 1987 ; O’Brien, 1993, 1998 ; Whittaker ., 2001 ), environmental heterogeneity ( Linder, 1991 ; Cowling ., 1997 ; Qian & Ricklefs, 2000 ), or both of these ( Clinebell ., 1995 ; O’Brien ., 2000 ). The mechanisms that underpin these relationships are numerous ( Fraser & Currie, 1996 ). There is strong support for explanations that favour contemporary ecological factors ( Wright, 1983 ; Currie & Paquin, 1987 ; O’Brien, 1998 ), as well as those that emphasize historical factors, which have produced differences in speciation and extinction rates ( Whittaker, 1977 ; Latham & Ricklefs, 1993 ; Ricklefs & Schluter, 1993 ; McGlone, 1996 ; Cowling ., 1997 ; Dynesius & Jansson, 2000 ; Qian & Ricklefs, 2000 ). When assessing this apparent tension between ‘ecological’ and ‘historical’ approaches, three things stand out: (i) at the scale of within‐biogeographical zones, steady state species diversity is primarily a function of speciation and extinction rates ( Rosenzweig, 1995 ); (ii) while these rates may be influenced by unpredictable and idiosyncratic events of the past (i.e. history), they are also governed by deterministic ecological patterns and processes ( Linder, 1985 ); and (iii) much of what is termed history, is — stated simply — the ecology of the past ( McGlone, 1996 ). South Africa’s Cape Floristic Region is very rich in plant species ( Goldblatt, 1997 ), most of which are the product of explosive speciation since the late Tertiary ( Richardson ., 2001 ). There has been much interest in describing and explaining plant diversity patterns in this region at the local ( Bond, 1983 ; Cowling, 1983 ; Wisheu ., 2000 ) and regional ( Kruger & Taylor, 1979 ; Cowling, 1990 ; Cowling ., 1992, 1997 ; Simmons & Cowling, 1996 ) scales. There is a well‐documented pattern at the regional scale, namely a concentration of species in the west, with a decline eastwards (e.g. Levyns, 1964 ; Oliver ., 1983 ; Cody, 1986 ). Cowling . (1992, 1997 ) elaborated this geographical pattern: they used species–area analysis to show that western (winter‐rainfall) landscapes (i.e. west of c . 21°E) had more than double the number of species than eastern (nonseasonal rainfall) landscapes across all area sizes. This pattern correlated with geographical differences in rainfall seasonality and reliability. Rainfall in the west is associated almost entirely with frontal systems during winter ( Deacon ., 1992 ). In the east, the situation is much more complex: here the eastwards penetration of frontal rains is restricted by the north–south trending axes of the Cape Folded Belt and most rain is associated with postfrontal events, especially in the spring and autumn. For any given rainfall total, western areas that receive predominantly frontal rain have a more reliable regime than in the east ( Cowling ., 1992 ). In addition to this geographical pattern, Linder (1991 ) observed a topographic pattern of regional plant diversity in the west. He showed that diversity in equal‐sized quarter‐degree squares (QDS) (634–671 km 2 ) was correlated significantly and positively with altitudinal and rainfall range (both measures of environmental heterogeneity), as well as with total rainfall, itself strongly related to rainfall range. Overall, richness was higher in the topographically and climatically more heterogeneous mountain than lowland landscapes. Interestingly, measures of environmental heterogeneity (topographical heterogeneity, rainfall and temperature range) and energy (potential evaporation, primary production, duration of growing season) did not emerge as significant determinants of this geographical pattern of regional diversity ( Cowling ., 1997 ). As stated above, this pattern correlates with patterns in rainfall seasonality and reliability. Cowling . (1997 ) found no significant differences for the sites they investigated (a subset of those in Appendix 1 ) in area, topographic diversity, length of rainfall gradient and length of temperature gradient. All these measures are very crude surrogates of environmental heterogeneity ( Scheiner, 1992 ). Cowling . (1992, 1997 ) postulated that the geographical pattern was a consequence of more pronounced climatic change during the Pleistocene in the east than the west, leading to different speciation and extinction histories. They argued that drier Pleistocene climates in the east reduced substantially the extent of Cape vegetation, especially species‐rich fynbos, thereby increasing extinction rates and disrupting the potential for speciation ( Dynesius & Jansson, 2000 ). In the west, however, similar or even slightly wetter‐than‐present Pleistocene climates enabled the persistence of Cape vegetation, thereby maintaining speciation processes and depressing extinction rates, at least relative to the east. Available palaeoecological data support this hypothesis (e.g. Meadows & Sugden, 1993 ; Scott ., 1997 ; Cowling ., 1999 ). Also cited as evidence in support of this hypothesis are the almost twofold higher levels of beta and gamma diversity ( sensu Cody, 1986 ) in western than eastern landscapes ( Cowling ., 1992 ; Simmons & Cowling, 1996 ) and the higher proportions of range‐restricted, habitat‐specialist species in western floras ( Cowling & McDonald, 1999 ). This pattern is suggestive of a higher tempo of speciation and/or greater persistence of range‐restricted beta (ecological specialist) and gamma (geographical vicariant) species in the west. Furthermore, phylogenetic data for several Cape lineages indicate the most massive and recent speciation has been in the west ( Schrire, 1991 ; Linder & Mann, 1998 ; Bakker ., 1999 ). There remain many unanswered questions on patterns of regional plant diversity across the Cape Floristic Region. For example, is the topographic pattern evident in the east; what is the role of environmental heterogeneity, measured as a biologically meaningful variable, in explaining geographical and topographic patterns; are there patterns in speciation/extinction histories and to what extent do these mirror diversity patterns; what is the role of contemporary climatic conditions, as opposed to historical ones, in explaining diversity patterns? We attempt to answer these questions in this paper. First, we report on geographical and topographic patterns of regional plant diversity. We then investigate the role of biological heterogeneity and speciation/extinction history as determinants of these patterns. We define biological heterogeneity as environmental variation that significantly influences the distribution and abundance of Cape plants. We used community diversity as a surrogate for biological heterogeneity because the number of communities in a landscape is likely to provide a more accurate measure of biologically significant environmental variation than crudely derived indices of topographical complexity and climatic range ( Scheiner, 1992 ). As an indicator of differences in speciation/extinction histories, we analysed patterns of rare species’ incidence across the region. We discuss our findings in relation to the evolution of plant diversity in the Cape Floristic Region. Methods Species diversity patterns Here we addressed the question: are there geographical and topographic patterns of regional plant diversity? First we divided the Cape Floristic Region into western (winter‐rainfall), eastern (non‐seasonal rainfall), lowland and montane subregions, following the method used by Cowling & Heijnis (2001 ) ( Fig. 1 ). This method involved separating strongly winter‐rainfall and winter‐rainfall homogeneous climate zones ( Schulze, 1997 ) from the rest (mainly non‐seasonal but also equinoctial regimes). Montane and lowland subregions were delineated along the boundary between broad habitat units in mountainous regions (mountain fynbos complex, most inland renosterveld types) and those on coastal lowlands, and interior valleys and basins. Broad habitat units are surrogates for vegetation type that are similar in geology, climate and topography ( Cowling & Heijnis, 2001 ). 1 The Cape Floristic Region showing the subregional classification used in this study. Next we compiled plant species–area data derived from sources cited in Cowling . (1992 ) and Cowling . (1997 ), as well from Taylor (1996 ) for the Cederberg ( Appendix 1 ). All datasets comprise complete, or near‐complete, inventories of species in a specified area, derived from collections, field notes and electronic databases. A total of 30 datasets were thus compiled, covering a wide range of richness and area values, and encompassing 32 of the 87 broad habitat units distributed across the Cape Floristic Region ( Appendix 1 ). Note that topographical heterogeneity, measured by Cowling . (1997 ) as the coefficient of variation of all the grid altitude values in an area (these being a subset of the sites in Appendix 1 ), showed no significant geographical variation within topographic subregions. Thus, for montane (west vs. east), t (unpaired with Welch correction) = 2.00, d.f. = 5, P = 0.06; and for lowland, t = 0.875, d.f. = 5, P = 0.42. The montane contrast is marginally non‐significant; in this case, the mean heterogeneity value was 2.3 times higher in the west than the east. We then fitted the species–area data to a double logarithmic regression model (log S = z log A + log c ) as this relationship is theoretically appropriate and provides the best fit for large areas within relatively homogeneous biogeographical zones ( Rosenzweig, 1995 ; see also Cowling ., 1992, 1997 ). Finally, we investigated geographical and topographic patterns in a number of ways. Using F ‐tests, we determined whether there were significant differences in the slopes ( Z ‐values) and intercepts ( c ‐values) between the following datasets: • western ( n = 18) vs. eastern ( n = 12); • montane ( n = 16) vs. lowland ( n = 14); • west‐montane ( n = 10) vs. west‐lowland ( n = 8); • east‐montane ( n = 6) vs. east‐lowland ( n = 6). Although we would not expect significant differences in Z ‐values between our different (mainland) datasets, differences in c ‐values are very important for comparing diversity patterns ( Rosenzweig, 1995 ). When the slopes of two curves are homogeneous, it is acceptable to compute the c ratio, or the ratio of the values of the intercepts in arithmetic space ( Gould, 1979 ). This ratio provides a measure of relative species densities when log A = 0 (i.e. 1 km 2 ). With increasing area, the curves diverge in arithmetic space as species accumulate at a faster for rate for the curve with the higher value of c ( Rosenzweig, 1995 ). We also used analysis of covariance — with area as the covariate — to analyse the geographical (western vs. eastern) and topographic (montane vs. lowland) patterns for the full dataset ( n = 30). Community diversity patterns We used the community diversity of an area as a surrogate for its biological heterogeneity or habitat diversity ( Rosenzweig, 1995 ). We did this because, in the Cape Floristic Region, plant species respond to very fine‐scale changes in soil and energy regimes ( Richards ., 1995 ; McDonald ., 1996 ) that cannot be quantified adequately using available, coarse‐scale climate, soils (geology) and topography databases. For example, the climatic data are available at a relatively crude scale (1′× 1′) ( Schulze, 1997 ) and comprise interpolations based on too few data sources for accurate patterns, especially in remote mountain landscapes. Furthermore, geology does not reflect accurately the diversity of soil types in a region ( Thwaites & Cowling, 1988 ). We reasoned that the actual diversity of plant communities in area would provide a measure of habitat diversity that is more appropriate as an explanatory variable for species diversity than the available physical measures of heterogeneity, such as topographic and climatic diversity ( Scheiner, 1992 ). The overall approach adopted for modelling community diversity patterns was similar to that for the species diversity analysis. Thus, we asked the question: is diversity uniform across the region? Again, we categorized sites as western and eastern (geographical pattern), and montane and lowland (topographic pattern) ( Fig. 1 ). We compiled data from the many plant community survey studies undertaken in the Cape Floristic Region ( Appendix 2 ). A total of 34 datasets were compiled, encompassing 28 of the 87 broad habitat units ( Cowling & Heijnis, 2001 ) in the region. Most of the studies that we used adopted the Zurich–Montpellier approach to phytosociological survey and analysis ( Werger, 1974 ). All used fixed relevé sizes (5 × 10 m or 10 × 10 m) and sampled the full range of community‐level variation in their respective study sites; and most used manual sorting of data matrices to identify communities based on character species. There were, however, differences in sampling approach that may have had a bearing on our results: while most studies sampled floristic variation across a demarcated and untransformed study site (e.g. a nature reserve or mountain catchment), others sampled along environmental gradients, and others were constrained by habitat loss to sampling only remnant vegetation. Given the relatively low sample size for each sampling approach, we could not determine whether there was any consistent bias, introduced by sampling approach, in relation to the geographical and topographic categories under investigation. We analysed the data in a number of steps. First, we established bivariate relationships between community number and sampling area in the study area (either provided or determined from maps of the respective study areas), as well as between this variable and relevé number, to provide a measure of sampling intensity. Intuitively, these variables are likely to be good predictors of community diversity. Being satisfied that these two variables were good predictors of community diversity, we used analysis of covariance — with area and plot number as covariates — to investigate a geographical (western vs. eastern) and topographic (montane vs. lowland) patterns. We tested the null hypothesis that there would be no geographical and topographic patterns of community diversity and, hence, biological heterogeneity (two‐tailed test). Rare plant diversity patterns We used patterns in the incidence of rare plants, specifically Red Data Book plants (updated from Hilton‐Taylor (1996 )), as an indicator of spatial differences in speciation/extinction history. The dataset is reasonably comprehensive in that the number of false absence records is regarded as acceptably low ( Lombard ., 1999 ). We assumed that rarity is associated with the early (post‐speciation) and late (pre‐extinction) phases of the taxon cycle ( Glazier, 1980 ; Chown, 1997 ; Rosenzweig & Lomolino, 1997 ). Therefore, areas with a high incidence of rare species have experienced higher rates of speciation and/or lower rates of extinction than areas with fewer rare species. The overall approach used for the analysis of rare plant patterns was similar to that for the species and community diversity analyses. Thus, we asked the question: are rare species uniformly distributed across the region? Again, we categorized sites as western and eastern (geographical pattern), and montane and lowland (topographic pattern) ( Fig. 1 ). We used a rare plant database comprising 1588 species and subspecies stored in 202 quarter‐degree squares (QDS) that intersect with the Cape Floristic Region. The database was modified for our study in the following way. First, we eliminated the QDS which had > 50% of their area outside of the study area, or in the ocean. This left us with 136 QDS. Such filtering is unlikely to have biased downwards scores of the affected QDS because their scores were not consistently lower than unaffected QDS ( Fig. 2 ). Secondly, we allocated the remaining QDS to one of the following subregions using a majority rule: west‐montane (29 QDS), west‐lowland (42), east‐montane (29), east‐lowland (34) ( Fig. 2 ). Thirdly, we edited the RDB list to include taxa categorized as: insufficiently known, indeterminate, rare, and rare/vulnerable; and to exclude taxa categorized as: vulnerable, vulnerable/endangered, endangered, and extinct. We reasoned that this would remove from our database those taxa whose rarity status owes largely to anthropogenic factors ( Hilton‐Taylor, 1996 ) such as habitat transformation and alien plant invasions ( Rebelo, 1992 ; Richardson ., 1996 ). Most rare plants in the Cape Floristric Region are so‐called naturally rare species (i.e. rarity is the product of ecological and evolutionary processes rather than human impacts) ( Trinder‐Smith ., 1996 ), although this is not the case for some lowland areas in the west ( Rebelo, 1992 ). After screening out the taxa whose rarity status was presumed to be human‐induced, we were left with 1034 taxa comprising 2658 records. Some 73% of these taxa have distributions restricted to three or less QDS with 35% confined to a single QDS; this suggests that most of our rare species sample are local endemics (see also Trinder‐Smith ., 1996 ). 2 The incidence of a subset of Red Data Book (RDB) taxa (species and subspecies) in quarter‐degree squares (QDS) (634–671 km 2 ) in the Cape Floristic Region. See text for details on filtering the RDB database and allocating QDS to subregions. Not all of the datasets for the different subregions were distributed normally ( P < 0.05; Kolmogorov–Smirnov test). Therefore, we used Mann–Whitney tests to investigate differences in the diversity of rare plants per quarter‐degree square across geographical and topographic subregions, and the Kruskal–Wallis test and Dunn’s multiple comparison test to investigate for differences across all four subregions. We tested the null hypothesis that there were no differences in the mean number of rare taxa per quarter‐degree square between the subregions (two‐tailed test). Because the rare plant database comprised data across the entire region (as opposed to the species and community data which comprised a limited number of sample sites), we also investigated the continuous patterns across the geographical gradient. We achieved this by plotting number of rare taxa per quarter‐degree square in relation to longitude and tested the significance of the relationship using Spearman’s correlation analysis. Results Species diversity patterns Area explained a high proportion of the variance (> 85%) in species diversity patterns except for the montane (46%) and east‐montane (64%) datasets ( Fig. 3 , Table 1 ). The intercept of the regression for western sites was significantly higher than that for the eastern ones. The c ratio for these two datasets was 2.11, indicating that western sites have slightly more than double the number of species than eastern sites at small area sizes. However, species accumulate faster with increasing area in western than eastern sites. 3 Plant species–area curves for (a) the entire Cape Floristic Region, (b) eastern (E) and western (W) subregions, and (c) lowland and montane subregions. 1 Results of species–area (log–log) regressions and comparisons of slopes and intercepts ( F ‐tests) for different subregions in the Cape Floristic Region. Subregion comparisons test for geographical (west vs. east), and topographic (montane vs. lowland) patterns of species diversity. Sig. = significance, SE = standard error, NA = not applicable (slopes not homogeneous) Subregion n r 2 Slope (SE) F ‐ratio (sig.) Intercept (SE) F ‐ratio (sig.) West 18 0.867 0.236(0.024) 2.554(0.038) East 12 0.846 0.274(0.035) 0.853(0.364) 2.230(0.079) 24.411(< 0.0001) Montane 16 0.462 0.138(0.040) 2.652(0.084) Lowland 14 0.876 0.258(0.028) 6.046(0.021) 2.383(0.047) NA West‐montane 10 0.857 0.187(0.022) 2.678(0.048) West‐lowland 8 0.933 0.245(0.040) 2.167(0.165) 2.468(0.040) 8.785 (0.010) East‐montane 6 0.640 0.225 0.084) 2.291(0.215) East‐lowland 6 0.945 0.319(0.034) 1.312(0.282) 2.213(0.067) 2.092 (0.179) There was a slight topography pattern imposed on the geographical pattern. For both montane and lowland subregions, the slopes of the regressions for western and eastern data were homogeneous ( F = 0.278, P = 0.607 and F = 2.659, P = 0.134, respectively) and the intercepts were significantly different ( F = 18.583, P = < 0.0001 and F = 7.622, P = 0.018, respectively). The c ratio values derived from these regressions were 2.43 for the montane data and 1.80 for the lowland data. Therefore, the west–east richness differential was 1.4 times higher in montane than in lowland areas. The slope of the species–area relationship for the montane sites was significantly lower than that for the lowland sites ( Table 1 , Fig. 3c ). Therefore, it was not permissible to compute the c ratios for these curves. Montane patterns were influenced strongly by the comparatively low richness of large‐sized eastern sites ( Fig. 3a ). There was a topographic pattern in the western subregion. As determined from the c ratio (slopes were homogeneous), western, montane sites of small size had 1.65 times as many species as comparable western, lowland sites. No such topographic pattern was evident in the east. Neither the slopes nor the intercepts of the eastern montane and lowland regressions were significantly different. The ancova results showed a highly significant geographical and a non‐significant topographic pattern ( Table 2 ). There was, however, a weakly significant interaction, a result of the fact that while western montane sites were richer than corresponding lowland sites, there was no difference between the richness of lowland and montane sites in the east. 2 Results of an analysis of covariance, with log area as the covariate, and plant species diversity as the response variable. Treatment factors investigate geographical (west vs. east), and topographic (montane vs. lowland) patterns Source of variation Sum of squares d.f. Mean squares F ratio Significance Covariates Log area 2.41 1 2.41 145.02 < 0.0001 Main effects (1) Geogr. 0.38 1 0.38 23.13 < 0.0005 (2) Topogr. 0.01 1 0.01 0.22 0.646 Interactions (1) × (2) 0.08 1 0.08 4.81 0.038 Residual 0.42 25 0.02 Total 3.42 29 (corrected) In summary, east‐montane sites were no richer than east‐lowland sites across all area sizes, yet in the west, montane sites were more than one‐and‐a‐half times richer than lowland sites. West‐montane sites were 2.5 times richer than montane sites in the east, while lowland sites in the west were about 1.8 times richer than corresponding sites in the east. Irrespective of topography, western sites were more than twice as rich as eastern sites. It is important to note that these values refer only to smaller‐sized areas; the differences will be larger as area is increased in a manner that is proportional to the differences in the c ‐values of the curves ( Rosenzweig, 1995 ). Community diversity patterns The best‐fit, linear relationships were achieved with the response variable (community diversity) untransformed and the explanatory variables log transformed (area) and untransformed (plot number). Both bivariate relationships were significant, although only a moderate proportion of the variance was explained ( Fig. 4 ). Given that the two explanatory variables were reasonable predictors of community diversity, their use as covariates in the ancova was justified. There were, however, neither significant geographical, nor significant topographic patterns of community diversity ( Table 3 ). The conclusion is that, unlike species diversity, community diversity (and, hence, biological heterogeneity) shows no significant variation in relation to locality in the four subregions. 4 Relationship between the diversity of phytosociologically recognized communities and size of the sampling domain, and sampling intensity in different subregions of the Cape Floristic Region. 3 Results of an analysis of covariance, with log area of the sampling domain and number of plots (or sampling intensity) as the covariates, and community diversity as the response variable. Treatment factors define geography (west vs. east) and topographic (montane vs. lowland) patterns Source of variation Sum of squares d.f. Mean squares F ratio Significance Covariates Log area 354.57 1 354.57 13.56 < 0.005 No. plots 262.27 1 262.27 10.03 < 0.005 Main effects (1) Geogr. 75.61 1 75.61 3.03 0.094 (2) Topogr. 44.75 1 44.75 1.79 0.190 Interactions (1) × (2) 56.07 1 56.07 2.25 0.150 Residual 623.62 25 24.95 Total 1542.71 30 (corrected) Rare plant diversity patterns The total number of species in each of the subregions was as follows: east‐lowland (241), east‐montane (246), west‐lowland (499), west‐montane (541). On average, there was double the number of rare plants per quarter‐degree square (QDS) in the western (25.6) than eastern (12.9) subregions ( Fig. 5a ). This difference was highly significant ( U = 1271, P < 0.0001). Lowland QDS had significantly fewer (16.7) rare plants than montane ones (23.2) ( U = 1799, P < 0.05) ( Fig. 5b ). However, this topographic pattern was evident only in the western subregion where the incidence or rare plants was 1.5 times higher in montane than lowland QDS ( Fig. 5c ). 5 Richness of a subset (naturally rare) of Red Data Book (RDB) taxa in quarter‐degree squares (QDS) (643–671 km 2 ) in different subregions of the Cape Floristic Region. Enclosed lines indicate the median, boxes the variance, and error bars the standard deviation. Abbreviations for subregions are as follows: E = east ( n = 63 QDS), W = west ( n = 71), L = lowland ( n = 76), M = montane ( n = 58 QDS). Diffferent superscripts indicate significant ( P < 0.05) differences between subregions using Mann–Whitney tests (Fig. 5a, b) and Dunn’s multiple comparison test (Fig. 5c). Maximum diversity of rare taxa per quarter‐degree square (QDS) in montane habitats declined abruptly between 18°30′ and 20°E, and thereafter gradually along the west–east geographical gradient ( Fig. 6 ). A generally similar pattern was evident on the lowlands, although some QDS in the extreme west had relatively low numbers of rare taxa. For both lowland and montane regions, the numbers of taxa per QDS were consistently low east of 23°E. 6 Relationship between richness and longitude for a subset (naturally rare) of Red Data Book (RDB) taxa in quarter‐degree squares (QDS) (634–671 km 2 ) in the Cape Floristic Region. Discussion Diversity patterns Our data confirm geographical and topographic (west only) patterns of regional plant diversity in the Cape Floristic Region, already described by Cowling . (1992 ) and Linder (1991 ), respectively. The geographical pattern, associated with a halving of diversity east of the divide, is remarkable. The boundary between the two regions coincides with the boundary between the winter and non‐seasonal rainfall zones. In actual fact, the zone of transition between these two rainfall regimes extends at least 100 km on either side of the boundary; there is probably a more gradual eastwards decline in diversity, as shown by patterns in quarter‐degree squares for individual taxa (e.g. Oliver ., 1983 ; Ojeda, 1998 ) — as well as rare species ( Fig. 6 ) — than an abrupt change along our somewhat arbitrary boundary. The arbitrary nature of the subregional delimitation, masked a possibly more subtle longitudinal diversity gradient. Interestingly, a topography pattern was evident only in the west. Linder (1991 ) argued that the greater climatic heterogeneity associated with montane landscapes was a more important determinant of plant diversity than geological heterogeneity, which is higher on the lowlands. However, despite no differences in measures of climatic and topographic heterogeneity between the eastern and western subregions (see Methods and Cowling ., 1997 ), and similar edaphic, climatic and topographic contrasts between montane and lowland landscapes in the two subregions ( Cowling ., 1992 ), a topographic pattern was not recorded in the east. Why do diversity patterns in the east differ from the west in such a pronounced way? Is this a result of different biological heterogeneities? It would seem not. Could it be, as suggested by Cowling . (1992 ), a consequence of different Pleistocene climates that have produced different vegetation and speciation histories? What about the role of the contemporary climate, specifically rainfall seasonality and reliability, in determining these patterns? These questions underpin much of the discussion that follows. The role of heterogeneity Many studies have shown that heterogeneity is an important predictor of plant diversity at the regional scale (e.g. Harner & Harper, 1976 ; Richerson & Lum, 1980 ; O’Brien ., 2000 ). This is also true of the Cape Floristic Region where a measure of topographic heterogeneity emerged as highly significant predictor of regional diversity ( Cowling ., 1997 ). However, our measure of biological heterogeneity — community diversity — showed no appreciable variation in relation to the subregions identified, even though species diversity did. What is the explanation? It appears that in the Cape Floristic Region there is no relationship between conventional measures of environmental heterogeneity (topographic diversity, climatic gradient length) and our measure of biological heterogeneity. Thus, lowland landscapes support as many plant communities as similar‐sized and more physiographically complex montane landscapes (cf. Linder, 1991 ). A possible explanation is that the higher edaphic complexity of lowland landscapes compensates for the lower topographic and climatic diversity as a determinant of biological heterogeneity. The high community diversity associated with edaphically diverse lowlands has been well documented in the west ( Boucher, 1987 ; Richards ., 1995 ) and the east ( Cowling, 1984 ). Clearly, Cape plants respond to measures of heterogeneity that are more subtle than the coarse variables used in most studies. A more difficult pattern to explain is the strong relationship between topographic heterogeneity and species diversity documented by Cowling . (1997 ). This relationship reflects the pattern of high diversity in the high‐heterogeneity landscapes of the western montane subregion, and the intermediate to low diversities of the low‐heterogeneity, lowland landscapes. As would be expected, the diversity of eastern, montane sites was consistently over‐predicted by the diversity–heterogeneity regression produced by Cowling . (1997 ). We suggest that biological heterogeneity per se is not the determinant of the high diversity in the western montane landscapes. Instead, we suggest that this relationship is a consequence of higher speciation rates — promoted by enhanced opportunities for isolation of populations — and lower extinction rates — associated with more refugia — in the topographically complex montane regions of the west; this would explain the very large numbers of rare species found there ( Simmons & Cowling, 1996 ). In the eastern mountains — as we hypothesize below — differences in contemporary and historical climatic regimes reduced rates of speciation and elevated extinction rates, leading to lower steady state diversities, irrespective of landscape ruggedness. The role of speciation/extinction history We have established that the western subregion supports more species than the east, and that in the west, the montane subregion supports more species than the lowland one. These patterns are not explained by differences in biological heterogeneity. We have also shown that patterns of rare plant diversity mirror overall diversity patterns. This suggests that regional diversity patterns are determined by differences in speciation and extinction rates ( Latham & Ricklefs, 1993 ; McGlone, 1996 ; Qian & Ricklefs, 2000 ). High diversity regions have large numbers of rare species, most of which are local endemics, which do not drive community‐level patterns. These are the species that ‘get the short bits of the broken stick’ ( Rosenzweig & Lomolino, 1997 ): they accumulate as entries at the bottom of phytosociological tables, explaining nothing about community patterns yet providing an answer to puzzling diversity patterns. These rare species are not a random subset of the flora of a region. They are significantly associated with a limited number of plant families (e.g. Ericaceae, Proteaceae, Rutaceae); and non‐sprouting (postfire) and limited gene dispersal (associated with short seed dispersal distance and insect pollination) are traits that are over‐represented among them ( Cowling & Holmes, 1992 ; McDonald ., 1995 ; Trinder‐Smith ., 1996 ). These biological traits, especially non‐sprouting, have favoured increased speciation rates and lower extinction rates. Thus, fire‐induced plant mortality increases generation turnover, thereby providing potential for more rapid evolution than sprouters ( Wisheu ., 2000 ). The higher allocation of resources by non‐sprouters to reproduction ( Bond & Midgley, 2001 ) increases their numeric dominance locally, and lowers extinction rates ( Wisheu ., 2000 ). Limited gene dispersal promotes isolation and hence speciation of daughter populations in marginal or unusual habitats: most rare species are habitat specialists ( Cowling ., 1992 ; Trinder‐Smith ., 1996 ). But why do speciation and extinction rates, and associated patterns of plant rarity and diversity, differ across the Cape Floristic Region? There are at least two hypotheses. The first is that originally stated by Cowling . (1992 ): the geographical pattern of diversity results from differences in Pleistocene climatic conditions and associated differences in the extent of Cape vegetation. In the west, fynbos and allied shrublands were not disrupted during wetter glacial periods; in the east, however, drier glacial conditions ( Deacon & Lancaster, 1988 ; Parkington ., 2000 ) probably restricted Cape vegetation to mesic refugia ( Cowling ., 1999 ). Not only did this result in the extinction of many species, it also disrupted speciation ( Dynesius & Jansson, 2000 ). Owing to the greater topographical diversity of mountain landscapes, more rare species persisted there than on the lowlands — hence, the topography pattern — albeit weak — in the west. Although there is some evidence of a higher incidence of range‐restricted rare species in mesic upland areas of the east ( Cowling & Campbell, 1983 ), we did not observe a topography pattern of rare and overall plant diversity there. The second hypothesis invokes contemporary climatic conditions, specifically gradients in rainfall seasonality and reliability that influence the success of different postfire regeneration biologies of Cape plants ( Ojeda, 1998 ). This we describe below. What about climate? Unlike in many other parts of the world (e.g. Richerson & Lum, 1980 ; Currie, 1991 ), neither mean annual rainfall nor available energy — measured as potential evaporation, primary production and duration of the growing season — explain diversity patterns in the Cape Floristic Region ( Cowling ., 1997 ;). The relatively low‐energy, low‐rainfall parts of the west support more species than higher‐energy, higher‐rainfall parts of the east (cf. O’Brien, 1993 ; O’Brien ., 2000 ). This is not to say that climate does not influence regional diversity patterns in the Cape Floristic Region. The geographical diversity pattern is associated with a shift from the predictable winter to the less predictable non‐seasonal rainfall zones of the region. Thus, multiple regressions incorporating area (which explained 60% of the variance in a simple regression) and a measure of rainfall concentration and rainfall reliability explained 79% and 84% of the variance in diversity, respectively ( Cowling ., 1997 ). The coefficients of variability of annual and seasonal rainfall in the west are significantly lower than in the east ( Cowling ., 1992 ; R.M. Cowling et al. , unpublished data). Predictable winter rainfall will favour non‐sprouting species as the survival of germinants would be enhanced by reliable rain after the summer fire season. Furthermore, non‐seasonal rainfall may favour sprouters — at least in small‐seeded taxa — as rainfall distributed throughout the year would facilitate the survival of seedlings that need to allocate large amounts of resources to below‐ground storage organs ( Ojeda, 1998 ). The higher incidence of sprouters in the eastern subregion ( Schutte ., 1995 ; Ojeda, 1998 ) may have negatively influenced the potential for speciation, owing to lower seedling production ( Bond & Midgley, 2001 ) and longer generation times ( Wisheu ., 2000 ). As a result of these differences in climate, and their influence on the success of different postfire regeneration biologies, speciation rates have been higher, and extinction rates lower, in the west than the east. Wisheu . (2000 ) suggest a soil nutrient explanation for the high numbers of non‐sprouters in Cape fynbos and Australian kwongan, both of which grow on impoverished sands, relative to other mediterranean‐climate vegetation that grows mainly on more fertile soils. In the former, nutrient‐poor environments, investment in underground organs is not worth the cost; non‐sprouters are favoured; and high speciation rates and low extinction rates lead to high steady‐state diversities. The patterns that we describe are consistent with this hypothesis as soil fertility decreases along lowland‐montane and east–west gradients in the Cape Floristic Region ( Campbell, 1983 ). However, rainfall reliability in the Cape (west only) and kwongan areas is significantly higher than in the other mediterranean‐climate areas (R.M. Cowling et al ., unpublished data); this could also explain the higher incidence of non‐sprouters — and higher diversity — in these soil‐impoverished regions. Conclusions We conclude by hypothesizing that regional diversity patterns in the Cape Floristic Region are the product of different speciation and extinction histories leading to different steady‐state diversities. There are, in fact, two hypotheses: (i) greater Pleistocene climatic stability in the west that would have resulted in higher rates of speciation and lower rates of extinction than in the east, where for the most, Pleistocene climates did not favour Cape lineages; and (ii) the more seasonal and reliable rainfall regime in the west would have favoured non‐sprouting plants and, hence, higher speciation rates and lower extinction rates, than in the east. Of these hypotheses, the second is more parsimonious, because poorly documented historical phenomena need not be invoked. However, drier glacial climates in the east, a consequence of low winter rain and little or no summer rain ( Cowling ., 1999 ), would mean that conditions there would have been even more unfavourable than now for nonsprouters throughout most of the Pleistocene. Both of these hypotheses are consistent with the higher incidence of rare species in the west, and higher levels of beta and gamma diversity there, associated with the turnover of these rare species — as well as locally common habitat specialist species ( Cowling, 1990 ) — along environmental and geographical gradients, respectively ( Cody, 1986 ; Cowling ., 1992 ). These rare species do not contribute to community patterns; hence, biological heterogeneity showed no geographical or topographic patterns. The weak topography pattern of diversity in the west arises from higher speciation rates and lower extinction rates in the topographically complex mountains, rather than from the influence of environmental heterogeneity on diversity. Appendix1 Species–area data for sites and associated broad habitat units (BHUs) ( Cowling & Heijnis, 2001 ) used in the analyses Broad habitat unit Area (km 2 ) No. spp. West‐lowland 0.15 157 Hagelkraal 0.15 157 Cape Flats 0.20 210 Boland 0.68 373 Cape Flats 1.37 229 Blackheath 2.70 379 Boland 10.40 585 De Hoop, Potberg, Agulhas 180.00 1179 Elim, Springfield, Hagelkraal, Agulhas 1609.25 1751 East‐lowland 0.67 150 Knysna 0.67 150 Aloes 1.41 217 St Francis 3.36 173 Goukamma 20.55 380 Suurbraak 27.86 446 Riversdale, Canca, Albertinia, Stilbaai, Gouritz, Blanco 2860.0 1580 West‐montane 0.27 364 Franschhoek 0.27 364 Kogelberg 1.58 533 Franschhoek 1.82 483 Klein River 6.02 697 Klein River 14.46 773 Franschhoek 45.30 1142 Cape Peninsula 77.50 1036 Kogelberg 240.0 1383 Cape Peninsula 471 2256 Cederberg 1259 1175 East montane 6.17 313 Outeniqua 6.17 313 Zuurberg 207.8 1100 Rooiberg 250.0 481 Klein Swartberg, Little Karoo 340.0 473 Southern Langeberg 1737.8 1203 Kouga, Baviaanskloof 1778.2 1122 Appendix 2 Plant community data for broad habitat units (BHU) (see Cowling & Heijnis, 2001 ) in the Cape Floristic Region Broad habitat unit Area sampled (km 2 ) No. relevés No. communities Data source West lowland Langebaan 250 157 24 Boucher (1987 ) Cape Flats 175 60 11 Boucher (1987 ) Hopefield 345 139 16 Boucher (1987 ) Blackheath 110 20 6 Boucher (1987 ) Springfield 0.15 42 3 Richards . (1995 ) Hagelkraal 0.15 33 2 Richards . (1995 ) Swartland 420 42 11 Boucher (1987 ) Boland 480 66 16 Boucher (1987 ) Overberg 49 23 3 Kemper (1997 ) East lowland 220 42 4 Cowling (1982 ) St Francis 220 42 4 Cowling (1982 ) Humansdorp 175 42 6 Cowling (1982 ) Kromme 375 27 3 Cowling (1982 ) West montane Cederberg 1260 197 26 Taylor (1996 ) Swartruggens 127 125 9 Lechmere‐Oertel (1998 ) Hawequas 9.7 105 10 Van Wilgen & Kruger (1985 ) Franschhoek 0.4 201 5 McDonald (1988 ) Cape Peninsula 2 48 3 Glyphis . (1978 ) 1.24 53 2 Joubert & Moll (1982 ) 1.4 38 2 Laidler . (1978 ) 4 78 18 McKenzie . (1977 ) 77 87 10 Privett (1998 ) Kogelberg 1.6 367 11 Kruger (1974 ) 240 250 29 Boucher (1978 ) East montane Southern Langeberg 1748 299 46 McDonald . (1996 ) Groot Swartberg 100 50 9 Bond (1981 ) Outeniqua 150 65 11 Bond (1981 ) Kouga 400 75 8 Euston‐Brown (1995 ) Baviaanskloof 200 38 6 Euston Brown (1995 ) Cockscomb 50 28 5 Cowling (1982 ) Acknowledgments This study was funded by the Global Environment Facility (through World Wide Fund–South Africa) and the University of Port Elizabeth. Mike Rosenzweig, Rob Whittaker and two anonymous referees made useful comments on an earlier draft. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Heterogeneity, speciation/extinction history and climate: explaining regional plant diversity patterns in the Cape Floristic Region

Download PDF
17 pages

Loading next page...
 
/lp/wiley/heterogeneity-speciation-extinction-history-and-climate-explaining-CtMPaWdP6e

References (89)

Publisher
Wiley
Copyright
Copyright © 2002 Wiley Subscription Services, Inc., A Wiley Company
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1046/j.1472-4642.2002.00143.x
Publisher site
See Article on Publisher Site

Abstract

Introduction Variation in plant diversity at the regional scale (0.1–10 6 km 2 ) within and between biogeographical zones is explained most frequently by models that incorporate as explanatory variables measures of climate/energy ( Richerson & Lum, 1980 ; Currie & Paquin, 1987 ; O’Brien, 1993, 1998 ; Whittaker ., 2001 ), environmental heterogeneity ( Linder, 1991 ; Cowling ., 1997 ; Qian & Ricklefs, 2000 ), or both of these ( Clinebell ., 1995 ; O’Brien ., 2000 ). The mechanisms that underpin these relationships are numerous ( Fraser & Currie, 1996 ). There is strong support for explanations that favour contemporary ecological factors ( Wright, 1983 ; Currie & Paquin, 1987 ; O’Brien, 1998 ), as well as those that emphasize historical factors, which have produced differences in speciation and extinction rates ( Whittaker, 1977 ; Latham & Ricklefs, 1993 ; Ricklefs & Schluter, 1993 ; McGlone, 1996 ; Cowling ., 1997 ; Dynesius & Jansson, 2000 ; Qian & Ricklefs, 2000 ). When assessing this apparent tension between ‘ecological’ and ‘historical’ approaches, three things stand out: (i) at the scale of within‐biogeographical zones, steady state species diversity is primarily a function of speciation and extinction rates ( Rosenzweig, 1995 ); (ii) while these rates may be influenced by unpredictable and idiosyncratic events of the past (i.e. history), they are also governed by deterministic ecological patterns and processes ( Linder, 1985 ); and (iii) much of what is termed history, is — stated simply — the ecology of the past ( McGlone, 1996 ). South Africa’s Cape Floristic Region is very rich in plant species ( Goldblatt, 1997 ), most of which are the product of explosive speciation since the late Tertiary ( Richardson ., 2001 ). There has been much interest in describing and explaining plant diversity patterns in this region at the local ( Bond, 1983 ; Cowling, 1983 ; Wisheu ., 2000 ) and regional ( Kruger & Taylor, 1979 ; Cowling, 1990 ; Cowling ., 1992, 1997 ; Simmons & Cowling, 1996 ) scales. There is a well‐documented pattern at the regional scale, namely a concentration of species in the west, with a decline eastwards (e.g. Levyns, 1964 ; Oliver ., 1983 ; Cody, 1986 ). Cowling . (1992, 1997 ) elaborated this geographical pattern: they used species–area analysis to show that western (winter‐rainfall) landscapes (i.e. west of c . 21°E) had more than double the number of species than eastern (nonseasonal rainfall) landscapes across all area sizes. This pattern correlated with geographical differences in rainfall seasonality and reliability. Rainfall in the west is associated almost entirely with frontal systems during winter ( Deacon ., 1992 ). In the east, the situation is much more complex: here the eastwards penetration of frontal rains is restricted by the north–south trending axes of the Cape Folded Belt and most rain is associated with postfrontal events, especially in the spring and autumn. For any given rainfall total, western areas that receive predominantly frontal rain have a more reliable regime than in the east ( Cowling ., 1992 ). In addition to this geographical pattern, Linder (1991 ) observed a topographic pattern of regional plant diversity in the west. He showed that diversity in equal‐sized quarter‐degree squares (QDS) (634–671 km 2 ) was correlated significantly and positively with altitudinal and rainfall range (both measures of environmental heterogeneity), as well as with total rainfall, itself strongly related to rainfall range. Overall, richness was higher in the topographically and climatically more heterogeneous mountain than lowland landscapes. Interestingly, measures of environmental heterogeneity (topographical heterogeneity, rainfall and temperature range) and energy (potential evaporation, primary production, duration of growing season) did not emerge as significant determinants of this geographical pattern of regional diversity ( Cowling ., 1997 ). As stated above, this pattern correlates with patterns in rainfall seasonality and reliability. Cowling . (1997 ) found no significant differences for the sites they investigated (a subset of those in Appendix 1 ) in area, topographic diversity, length of rainfall gradient and length of temperature gradient. All these measures are very crude surrogates of environmental heterogeneity ( Scheiner, 1992 ). Cowling . (1992, 1997 ) postulated that the geographical pattern was a consequence of more pronounced climatic change during the Pleistocene in the east than the west, leading to different speciation and extinction histories. They argued that drier Pleistocene climates in the east reduced substantially the extent of Cape vegetation, especially species‐rich fynbos, thereby increasing extinction rates and disrupting the potential for speciation ( Dynesius & Jansson, 2000 ). In the west, however, similar or even slightly wetter‐than‐present Pleistocene climates enabled the persistence of Cape vegetation, thereby maintaining speciation processes and depressing extinction rates, at least relative to the east. Available palaeoecological data support this hypothesis (e.g. Meadows & Sugden, 1993 ; Scott ., 1997 ; Cowling ., 1999 ). Also cited as evidence in support of this hypothesis are the almost twofold higher levels of beta and gamma diversity ( sensu Cody, 1986 ) in western than eastern landscapes ( Cowling ., 1992 ; Simmons & Cowling, 1996 ) and the higher proportions of range‐restricted, habitat‐specialist species in western floras ( Cowling & McDonald, 1999 ). This pattern is suggestive of a higher tempo of speciation and/or greater persistence of range‐restricted beta (ecological specialist) and gamma (geographical vicariant) species in the west. Furthermore, phylogenetic data for several Cape lineages indicate the most massive and recent speciation has been in the west ( Schrire, 1991 ; Linder & Mann, 1998 ; Bakker ., 1999 ). There remain many unanswered questions on patterns of regional plant diversity across the Cape Floristic Region. For example, is the topographic pattern evident in the east; what is the role of environmental heterogeneity, measured as a biologically meaningful variable, in explaining geographical and topographic patterns; are there patterns in speciation/extinction histories and to what extent do these mirror diversity patterns; what is the role of contemporary climatic conditions, as opposed to historical ones, in explaining diversity patterns? We attempt to answer these questions in this paper. First, we report on geographical and topographic patterns of regional plant diversity. We then investigate the role of biological heterogeneity and speciation/extinction history as determinants of these patterns. We define biological heterogeneity as environmental variation that significantly influences the distribution and abundance of Cape plants. We used community diversity as a surrogate for biological heterogeneity because the number of communities in a landscape is likely to provide a more accurate measure of biologically significant environmental variation than crudely derived indices of topographical complexity and climatic range ( Scheiner, 1992 ). As an indicator of differences in speciation/extinction histories, we analysed patterns of rare species’ incidence across the region. We discuss our findings in relation to the evolution of plant diversity in the Cape Floristic Region. Methods Species diversity patterns Here we addressed the question: are there geographical and topographic patterns of regional plant diversity? First we divided the Cape Floristic Region into western (winter‐rainfall), eastern (non‐seasonal rainfall), lowland and montane subregions, following the method used by Cowling & Heijnis (2001 ) ( Fig. 1 ). This method involved separating strongly winter‐rainfall and winter‐rainfall homogeneous climate zones ( Schulze, 1997 ) from the rest (mainly non‐seasonal but also equinoctial regimes). Montane and lowland subregions were delineated along the boundary between broad habitat units in mountainous regions (mountain fynbos complex, most inland renosterveld types) and those on coastal lowlands, and interior valleys and basins. Broad habitat units are surrogates for vegetation type that are similar in geology, climate and topography ( Cowling & Heijnis, 2001 ). 1 The Cape Floristic Region showing the subregional classification used in this study. Next we compiled plant species–area data derived from sources cited in Cowling . (1992 ) and Cowling . (1997 ), as well from Taylor (1996 ) for the Cederberg ( Appendix 1 ). All datasets comprise complete, or near‐complete, inventories of species in a specified area, derived from collections, field notes and electronic databases. A total of 30 datasets were thus compiled, covering a wide range of richness and area values, and encompassing 32 of the 87 broad habitat units distributed across the Cape Floristic Region ( Appendix 1 ). Note that topographical heterogeneity, measured by Cowling . (1997 ) as the coefficient of variation of all the grid altitude values in an area (these being a subset of the sites in Appendix 1 ), showed no significant geographical variation within topographic subregions. Thus, for montane (west vs. east), t (unpaired with Welch correction) = 2.00, d.f. = 5, P = 0.06; and for lowland, t = 0.875, d.f. = 5, P = 0.42. The montane contrast is marginally non‐significant; in this case, the mean heterogeneity value was 2.3 times higher in the west than the east. We then fitted the species–area data to a double logarithmic regression model (log S = z log A + log c ) as this relationship is theoretically appropriate and provides the best fit for large areas within relatively homogeneous biogeographical zones ( Rosenzweig, 1995 ; see also Cowling ., 1992, 1997 ). Finally, we investigated geographical and topographic patterns in a number of ways. Using F ‐tests, we determined whether there were significant differences in the slopes ( Z ‐values) and intercepts ( c ‐values) between the following datasets: • western ( n = 18) vs. eastern ( n = 12); • montane ( n = 16) vs. lowland ( n = 14); • west‐montane ( n = 10) vs. west‐lowland ( n = 8); • east‐montane ( n = 6) vs. east‐lowland ( n = 6). Although we would not expect significant differences in Z ‐values between our different (mainland) datasets, differences in c ‐values are very important for comparing diversity patterns ( Rosenzweig, 1995 ). When the slopes of two curves are homogeneous, it is acceptable to compute the c ratio, or the ratio of the values of the intercepts in arithmetic space ( Gould, 1979 ). This ratio provides a measure of relative species densities when log A = 0 (i.e. 1 km 2 ). With increasing area, the curves diverge in arithmetic space as species accumulate at a faster for rate for the curve with the higher value of c ( Rosenzweig, 1995 ). We also used analysis of covariance — with area as the covariate — to analyse the geographical (western vs. eastern) and topographic (montane vs. lowland) patterns for the full dataset ( n = 30). Community diversity patterns We used the community diversity of an area as a surrogate for its biological heterogeneity or habitat diversity ( Rosenzweig, 1995 ). We did this because, in the Cape Floristic Region, plant species respond to very fine‐scale changes in soil and energy regimes ( Richards ., 1995 ; McDonald ., 1996 ) that cannot be quantified adequately using available, coarse‐scale climate, soils (geology) and topography databases. For example, the climatic data are available at a relatively crude scale (1′× 1′) ( Schulze, 1997 ) and comprise interpolations based on too few data sources for accurate patterns, especially in remote mountain landscapes. Furthermore, geology does not reflect accurately the diversity of soil types in a region ( Thwaites & Cowling, 1988 ). We reasoned that the actual diversity of plant communities in area would provide a measure of habitat diversity that is more appropriate as an explanatory variable for species diversity than the available physical measures of heterogeneity, such as topographic and climatic diversity ( Scheiner, 1992 ). The overall approach adopted for modelling community diversity patterns was similar to that for the species diversity analysis. Thus, we asked the question: is diversity uniform across the region? Again, we categorized sites as western and eastern (geographical pattern), and montane and lowland (topographic pattern) ( Fig. 1 ). We compiled data from the many plant community survey studies undertaken in the Cape Floristic Region ( Appendix 2 ). A total of 34 datasets were compiled, encompassing 28 of the 87 broad habitat units ( Cowling & Heijnis, 2001 ) in the region. Most of the studies that we used adopted the Zurich–Montpellier approach to phytosociological survey and analysis ( Werger, 1974 ). All used fixed relevé sizes (5 × 10 m or 10 × 10 m) and sampled the full range of community‐level variation in their respective study sites; and most used manual sorting of data matrices to identify communities based on character species. There were, however, differences in sampling approach that may have had a bearing on our results: while most studies sampled floristic variation across a demarcated and untransformed study site (e.g. a nature reserve or mountain catchment), others sampled along environmental gradients, and others were constrained by habitat loss to sampling only remnant vegetation. Given the relatively low sample size for each sampling approach, we could not determine whether there was any consistent bias, introduced by sampling approach, in relation to the geographical and topographic categories under investigation. We analysed the data in a number of steps. First, we established bivariate relationships between community number and sampling area in the study area (either provided or determined from maps of the respective study areas), as well as between this variable and relevé number, to provide a measure of sampling intensity. Intuitively, these variables are likely to be good predictors of community diversity. Being satisfied that these two variables were good predictors of community diversity, we used analysis of covariance — with area and plot number as covariates — to investigate a geographical (western vs. eastern) and topographic (montane vs. lowland) patterns. We tested the null hypothesis that there would be no geographical and topographic patterns of community diversity and, hence, biological heterogeneity (two‐tailed test). Rare plant diversity patterns We used patterns in the incidence of rare plants, specifically Red Data Book plants (updated from Hilton‐Taylor (1996 )), as an indicator of spatial differences in speciation/extinction history. The dataset is reasonably comprehensive in that the number of false absence records is regarded as acceptably low ( Lombard ., 1999 ). We assumed that rarity is associated with the early (post‐speciation) and late (pre‐extinction) phases of the taxon cycle ( Glazier, 1980 ; Chown, 1997 ; Rosenzweig & Lomolino, 1997 ). Therefore, areas with a high incidence of rare species have experienced higher rates of speciation and/or lower rates of extinction than areas with fewer rare species. The overall approach used for the analysis of rare plant patterns was similar to that for the species and community diversity analyses. Thus, we asked the question: are rare species uniformly distributed across the region? Again, we categorized sites as western and eastern (geographical pattern), and montane and lowland (topographic pattern) ( Fig. 1 ). We used a rare plant database comprising 1588 species and subspecies stored in 202 quarter‐degree squares (QDS) that intersect with the Cape Floristic Region. The database was modified for our study in the following way. First, we eliminated the QDS which had > 50% of their area outside of the study area, or in the ocean. This left us with 136 QDS. Such filtering is unlikely to have biased downwards scores of the affected QDS because their scores were not consistently lower than unaffected QDS ( Fig. 2 ). Secondly, we allocated the remaining QDS to one of the following subregions using a majority rule: west‐montane (29 QDS), west‐lowland (42), east‐montane (29), east‐lowland (34) ( Fig. 2 ). Thirdly, we edited the RDB list to include taxa categorized as: insufficiently known, indeterminate, rare, and rare/vulnerable; and to exclude taxa categorized as: vulnerable, vulnerable/endangered, endangered, and extinct. We reasoned that this would remove from our database those taxa whose rarity status owes largely to anthropogenic factors ( Hilton‐Taylor, 1996 ) such as habitat transformation and alien plant invasions ( Rebelo, 1992 ; Richardson ., 1996 ). Most rare plants in the Cape Floristric Region are so‐called naturally rare species (i.e. rarity is the product of ecological and evolutionary processes rather than human impacts) ( Trinder‐Smith ., 1996 ), although this is not the case for some lowland areas in the west ( Rebelo, 1992 ). After screening out the taxa whose rarity status was presumed to be human‐induced, we were left with 1034 taxa comprising 2658 records. Some 73% of these taxa have distributions restricted to three or less QDS with 35% confined to a single QDS; this suggests that most of our rare species sample are local endemics (see also Trinder‐Smith ., 1996 ). 2 The incidence of a subset of Red Data Book (RDB) taxa (species and subspecies) in quarter‐degree squares (QDS) (634–671 km 2 ) in the Cape Floristic Region. See text for details on filtering the RDB database and allocating QDS to subregions. Not all of the datasets for the different subregions were distributed normally ( P < 0.05; Kolmogorov–Smirnov test). Therefore, we used Mann–Whitney tests to investigate differences in the diversity of rare plants per quarter‐degree square across geographical and topographic subregions, and the Kruskal–Wallis test and Dunn’s multiple comparison test to investigate for differences across all four subregions. We tested the null hypothesis that there were no differences in the mean number of rare taxa per quarter‐degree square between the subregions (two‐tailed test). Because the rare plant database comprised data across the entire region (as opposed to the species and community data which comprised a limited number of sample sites), we also investigated the continuous patterns across the geographical gradient. We achieved this by plotting number of rare taxa per quarter‐degree square in relation to longitude and tested the significance of the relationship using Spearman’s correlation analysis. Results Species diversity patterns Area explained a high proportion of the variance (> 85%) in species diversity patterns except for the montane (46%) and east‐montane (64%) datasets ( Fig. 3 , Table 1 ). The intercept of the regression for western sites was significantly higher than that for the eastern ones. The c ratio for these two datasets was 2.11, indicating that western sites have slightly more than double the number of species than eastern sites at small area sizes. However, species accumulate faster with increasing area in western than eastern sites. 3 Plant species–area curves for (a) the entire Cape Floristic Region, (b) eastern (E) and western (W) subregions, and (c) lowland and montane subregions. 1 Results of species–area (log–log) regressions and comparisons of slopes and intercepts ( F ‐tests) for different subregions in the Cape Floristic Region. Subregion comparisons test for geographical (west vs. east), and topographic (montane vs. lowland) patterns of species diversity. Sig. = significance, SE = standard error, NA = not applicable (slopes not homogeneous) Subregion n r 2 Slope (SE) F ‐ratio (sig.) Intercept (SE) F ‐ratio (sig.) West 18 0.867 0.236(0.024) 2.554(0.038) East 12 0.846 0.274(0.035) 0.853(0.364) 2.230(0.079) 24.411(< 0.0001) Montane 16 0.462 0.138(0.040) 2.652(0.084) Lowland 14 0.876 0.258(0.028) 6.046(0.021) 2.383(0.047) NA West‐montane 10 0.857 0.187(0.022) 2.678(0.048) West‐lowland 8 0.933 0.245(0.040) 2.167(0.165) 2.468(0.040) 8.785 (0.010) East‐montane 6 0.640 0.225 0.084) 2.291(0.215) East‐lowland 6 0.945 0.319(0.034) 1.312(0.282) 2.213(0.067) 2.092 (0.179) There was a slight topography pattern imposed on the geographical pattern. For both montane and lowland subregions, the slopes of the regressions for western and eastern data were homogeneous ( F = 0.278, P = 0.607 and F = 2.659, P = 0.134, respectively) and the intercepts were significantly different ( F = 18.583, P = < 0.0001 and F = 7.622, P = 0.018, respectively). The c ratio values derived from these regressions were 2.43 for the montane data and 1.80 for the lowland data. Therefore, the west–east richness differential was 1.4 times higher in montane than in lowland areas. The slope of the species–area relationship for the montane sites was significantly lower than that for the lowland sites ( Table 1 , Fig. 3c ). Therefore, it was not permissible to compute the c ratios for these curves. Montane patterns were influenced strongly by the comparatively low richness of large‐sized eastern sites ( Fig. 3a ). There was a topographic pattern in the western subregion. As determined from the c ratio (slopes were homogeneous), western, montane sites of small size had 1.65 times as many species as comparable western, lowland sites. No such topographic pattern was evident in the east. Neither the slopes nor the intercepts of the eastern montane and lowland regressions were significantly different. The ancova results showed a highly significant geographical and a non‐significant topographic pattern ( Table 2 ). There was, however, a weakly significant interaction, a result of the fact that while western montane sites were richer than corresponding lowland sites, there was no difference between the richness of lowland and montane sites in the east. 2 Results of an analysis of covariance, with log area as the covariate, and plant species diversity as the response variable. Treatment factors investigate geographical (west vs. east), and topographic (montane vs. lowland) patterns Source of variation Sum of squares d.f. Mean squares F ratio Significance Covariates Log area 2.41 1 2.41 145.02 < 0.0001 Main effects (1) Geogr. 0.38 1 0.38 23.13 < 0.0005 (2) Topogr. 0.01 1 0.01 0.22 0.646 Interactions (1) × (2) 0.08 1 0.08 4.81 0.038 Residual 0.42 25 0.02 Total 3.42 29 (corrected) In summary, east‐montane sites were no richer than east‐lowland sites across all area sizes, yet in the west, montane sites were more than one‐and‐a‐half times richer than lowland sites. West‐montane sites were 2.5 times richer than montane sites in the east, while lowland sites in the west were about 1.8 times richer than corresponding sites in the east. Irrespective of topography, western sites were more than twice as rich as eastern sites. It is important to note that these values refer only to smaller‐sized areas; the differences will be larger as area is increased in a manner that is proportional to the differences in the c ‐values of the curves ( Rosenzweig, 1995 ). Community diversity patterns The best‐fit, linear relationships were achieved with the response variable (community diversity) untransformed and the explanatory variables log transformed (area) and untransformed (plot number). Both bivariate relationships were significant, although only a moderate proportion of the variance was explained ( Fig. 4 ). Given that the two explanatory variables were reasonable predictors of community diversity, their use as covariates in the ancova was justified. There were, however, neither significant geographical, nor significant topographic patterns of community diversity ( Table 3 ). The conclusion is that, unlike species diversity, community diversity (and, hence, biological heterogeneity) shows no significant variation in relation to locality in the four subregions. 4 Relationship between the diversity of phytosociologically recognized communities and size of the sampling domain, and sampling intensity in different subregions of the Cape Floristic Region. 3 Results of an analysis of covariance, with log area of the sampling domain and number of plots (or sampling intensity) as the covariates, and community diversity as the response variable. Treatment factors define geography (west vs. east) and topographic (montane vs. lowland) patterns Source of variation Sum of squares d.f. Mean squares F ratio Significance Covariates Log area 354.57 1 354.57 13.56 < 0.005 No. plots 262.27 1 262.27 10.03 < 0.005 Main effects (1) Geogr. 75.61 1 75.61 3.03 0.094 (2) Topogr. 44.75 1 44.75 1.79 0.190 Interactions (1) × (2) 56.07 1 56.07 2.25 0.150 Residual 623.62 25 24.95 Total 1542.71 30 (corrected) Rare plant diversity patterns The total number of species in each of the subregions was as follows: east‐lowland (241), east‐montane (246), west‐lowland (499), west‐montane (541). On average, there was double the number of rare plants per quarter‐degree square (QDS) in the western (25.6) than eastern (12.9) subregions ( Fig. 5a ). This difference was highly significant ( U = 1271, P < 0.0001). Lowland QDS had significantly fewer (16.7) rare plants than montane ones (23.2) ( U = 1799, P < 0.05) ( Fig. 5b ). However, this topographic pattern was evident only in the western subregion where the incidence or rare plants was 1.5 times higher in montane than lowland QDS ( Fig. 5c ). 5 Richness of a subset (naturally rare) of Red Data Book (RDB) taxa in quarter‐degree squares (QDS) (643–671 km 2 ) in different subregions of the Cape Floristic Region. Enclosed lines indicate the median, boxes the variance, and error bars the standard deviation. Abbreviations for subregions are as follows: E = east ( n = 63 QDS), W = west ( n = 71), L = lowland ( n = 76), M = montane ( n = 58 QDS). Diffferent superscripts indicate significant ( P < 0.05) differences between subregions using Mann–Whitney tests (Fig. 5a, b) and Dunn’s multiple comparison test (Fig. 5c). Maximum diversity of rare taxa per quarter‐degree square (QDS) in montane habitats declined abruptly between 18°30′ and 20°E, and thereafter gradually along the west–east geographical gradient ( Fig. 6 ). A generally similar pattern was evident on the lowlands, although some QDS in the extreme west had relatively low numbers of rare taxa. For both lowland and montane regions, the numbers of taxa per QDS were consistently low east of 23°E. 6 Relationship between richness and longitude for a subset (naturally rare) of Red Data Book (RDB) taxa in quarter‐degree squares (QDS) (634–671 km 2 ) in the Cape Floristic Region. Discussion Diversity patterns Our data confirm geographical and topographic (west only) patterns of regional plant diversity in the Cape Floristic Region, already described by Cowling . (1992 ) and Linder (1991 ), respectively. The geographical pattern, associated with a halving of diversity east of the divide, is remarkable. The boundary between the two regions coincides with the boundary between the winter and non‐seasonal rainfall zones. In actual fact, the zone of transition between these two rainfall regimes extends at least 100 km on either side of the boundary; there is probably a more gradual eastwards decline in diversity, as shown by patterns in quarter‐degree squares for individual taxa (e.g. Oliver ., 1983 ; Ojeda, 1998 ) — as well as rare species ( Fig. 6 ) — than an abrupt change along our somewhat arbitrary boundary. The arbitrary nature of the subregional delimitation, masked a possibly more subtle longitudinal diversity gradient. Interestingly, a topography pattern was evident only in the west. Linder (1991 ) argued that the greater climatic heterogeneity associated with montane landscapes was a more important determinant of plant diversity than geological heterogeneity, which is higher on the lowlands. However, despite no differences in measures of climatic and topographic heterogeneity between the eastern and western subregions (see Methods and Cowling ., 1997 ), and similar edaphic, climatic and topographic contrasts between montane and lowland landscapes in the two subregions ( Cowling ., 1992 ), a topographic pattern was not recorded in the east. Why do diversity patterns in the east differ from the west in such a pronounced way? Is this a result of different biological heterogeneities? It would seem not. Could it be, as suggested by Cowling . (1992 ), a consequence of different Pleistocene climates that have produced different vegetation and speciation histories? What about the role of the contemporary climate, specifically rainfall seasonality and reliability, in determining these patterns? These questions underpin much of the discussion that follows. The role of heterogeneity Many studies have shown that heterogeneity is an important predictor of plant diversity at the regional scale (e.g. Harner & Harper, 1976 ; Richerson & Lum, 1980 ; O’Brien ., 2000 ). This is also true of the Cape Floristic Region where a measure of topographic heterogeneity emerged as highly significant predictor of regional diversity ( Cowling ., 1997 ). However, our measure of biological heterogeneity — community diversity — showed no appreciable variation in relation to the subregions identified, even though species diversity did. What is the explanation? It appears that in the Cape Floristic Region there is no relationship between conventional measures of environmental heterogeneity (topographic diversity, climatic gradient length) and our measure of biological heterogeneity. Thus, lowland landscapes support as many plant communities as similar‐sized and more physiographically complex montane landscapes (cf. Linder, 1991 ). A possible explanation is that the higher edaphic complexity of lowland landscapes compensates for the lower topographic and climatic diversity as a determinant of biological heterogeneity. The high community diversity associated with edaphically diverse lowlands has been well documented in the west ( Boucher, 1987 ; Richards ., 1995 ) and the east ( Cowling, 1984 ). Clearly, Cape plants respond to measures of heterogeneity that are more subtle than the coarse variables used in most studies. A more difficult pattern to explain is the strong relationship between topographic heterogeneity and species diversity documented by Cowling . (1997 ). This relationship reflects the pattern of high diversity in the high‐heterogeneity landscapes of the western montane subregion, and the intermediate to low diversities of the low‐heterogeneity, lowland landscapes. As would be expected, the diversity of eastern, montane sites was consistently over‐predicted by the diversity–heterogeneity regression produced by Cowling . (1997 ). We suggest that biological heterogeneity per se is not the determinant of the high diversity in the western montane landscapes. Instead, we suggest that this relationship is a consequence of higher speciation rates — promoted by enhanced opportunities for isolation of populations — and lower extinction rates — associated with more refugia — in the topographically complex montane regions of the west; this would explain the very large numbers of rare species found there ( Simmons & Cowling, 1996 ). In the eastern mountains — as we hypothesize below — differences in contemporary and historical climatic regimes reduced rates of speciation and elevated extinction rates, leading to lower steady state diversities, irrespective of landscape ruggedness. The role of speciation/extinction history We have established that the western subregion supports more species than the east, and that in the west, the montane subregion supports more species than the lowland one. These patterns are not explained by differences in biological heterogeneity. We have also shown that patterns of rare plant diversity mirror overall diversity patterns. This suggests that regional diversity patterns are determined by differences in speciation and extinction rates ( Latham & Ricklefs, 1993 ; McGlone, 1996 ; Qian & Ricklefs, 2000 ). High diversity regions have large numbers of rare species, most of which are local endemics, which do not drive community‐level patterns. These are the species that ‘get the short bits of the broken stick’ ( Rosenzweig & Lomolino, 1997 ): they accumulate as entries at the bottom of phytosociological tables, explaining nothing about community patterns yet providing an answer to puzzling diversity patterns. These rare species are not a random subset of the flora of a region. They are significantly associated with a limited number of plant families (e.g. Ericaceae, Proteaceae, Rutaceae); and non‐sprouting (postfire) and limited gene dispersal (associated with short seed dispersal distance and insect pollination) are traits that are over‐represented among them ( Cowling & Holmes, 1992 ; McDonald ., 1995 ; Trinder‐Smith ., 1996 ). These biological traits, especially non‐sprouting, have favoured increased speciation rates and lower extinction rates. Thus, fire‐induced plant mortality increases generation turnover, thereby providing potential for more rapid evolution than sprouters ( Wisheu ., 2000 ). The higher allocation of resources by non‐sprouters to reproduction ( Bond & Midgley, 2001 ) increases their numeric dominance locally, and lowers extinction rates ( Wisheu ., 2000 ). Limited gene dispersal promotes isolation and hence speciation of daughter populations in marginal or unusual habitats: most rare species are habitat specialists ( Cowling ., 1992 ; Trinder‐Smith ., 1996 ). But why do speciation and extinction rates, and associated patterns of plant rarity and diversity, differ across the Cape Floristic Region? There are at least two hypotheses. The first is that originally stated by Cowling . (1992 ): the geographical pattern of diversity results from differences in Pleistocene climatic conditions and associated differences in the extent of Cape vegetation. In the west, fynbos and allied shrublands were not disrupted during wetter glacial periods; in the east, however, drier glacial conditions ( Deacon & Lancaster, 1988 ; Parkington ., 2000 ) probably restricted Cape vegetation to mesic refugia ( Cowling ., 1999 ). Not only did this result in the extinction of many species, it also disrupted speciation ( Dynesius & Jansson, 2000 ). Owing to the greater topographical diversity of mountain landscapes, more rare species persisted there than on the lowlands — hence, the topography pattern — albeit weak — in the west. Although there is some evidence of a higher incidence of range‐restricted rare species in mesic upland areas of the east ( Cowling & Campbell, 1983 ), we did not observe a topography pattern of rare and overall plant diversity there. The second hypothesis invokes contemporary climatic conditions, specifically gradients in rainfall seasonality and reliability that influence the success of different postfire regeneration biologies of Cape plants ( Ojeda, 1998 ). This we describe below. What about climate? Unlike in many other parts of the world (e.g. Richerson & Lum, 1980 ; Currie, 1991 ), neither mean annual rainfall nor available energy — measured as potential evaporation, primary production and duration of the growing season — explain diversity patterns in the Cape Floristic Region ( Cowling ., 1997 ;). The relatively low‐energy, low‐rainfall parts of the west support more species than higher‐energy, higher‐rainfall parts of the east (cf. O’Brien, 1993 ; O’Brien ., 2000 ). This is not to say that climate does not influence regional diversity patterns in the Cape Floristic Region. The geographical diversity pattern is associated with a shift from the predictable winter to the less predictable non‐seasonal rainfall zones of the region. Thus, multiple regressions incorporating area (which explained 60% of the variance in a simple regression) and a measure of rainfall concentration and rainfall reliability explained 79% and 84% of the variance in diversity, respectively ( Cowling ., 1997 ). The coefficients of variability of annual and seasonal rainfall in the west are significantly lower than in the east ( Cowling ., 1992 ; R.M. Cowling et al. , unpublished data). Predictable winter rainfall will favour non‐sprouting species as the survival of germinants would be enhanced by reliable rain after the summer fire season. Furthermore, non‐seasonal rainfall may favour sprouters — at least in small‐seeded taxa — as rainfall distributed throughout the year would facilitate the survival of seedlings that need to allocate large amounts of resources to below‐ground storage organs ( Ojeda, 1998 ). The higher incidence of sprouters in the eastern subregion ( Schutte ., 1995 ; Ojeda, 1998 ) may have negatively influenced the potential for speciation, owing to lower seedling production ( Bond & Midgley, 2001 ) and longer generation times ( Wisheu ., 2000 ). As a result of these differences in climate, and their influence on the success of different postfire regeneration biologies, speciation rates have been higher, and extinction rates lower, in the west than the east. Wisheu . (2000 ) suggest a soil nutrient explanation for the high numbers of non‐sprouters in Cape fynbos and Australian kwongan, both of which grow on impoverished sands, relative to other mediterranean‐climate vegetation that grows mainly on more fertile soils. In the former, nutrient‐poor environments, investment in underground organs is not worth the cost; non‐sprouters are favoured; and high speciation rates and low extinction rates lead to high steady‐state diversities. The patterns that we describe are consistent with this hypothesis as soil fertility decreases along lowland‐montane and east–west gradients in the Cape Floristic Region ( Campbell, 1983 ). However, rainfall reliability in the Cape (west only) and kwongan areas is significantly higher than in the other mediterranean‐climate areas (R.M. Cowling et al ., unpublished data); this could also explain the higher incidence of non‐sprouters — and higher diversity — in these soil‐impoverished regions. Conclusions We conclude by hypothesizing that regional diversity patterns in the Cape Floristic Region are the product of different speciation and extinction histories leading to different steady‐state diversities. There are, in fact, two hypotheses: (i) greater Pleistocene climatic stability in the west that would have resulted in higher rates of speciation and lower rates of extinction than in the east, where for the most, Pleistocene climates did not favour Cape lineages; and (ii) the more seasonal and reliable rainfall regime in the west would have favoured non‐sprouting plants and, hence, higher speciation rates and lower extinction rates, than in the east. Of these hypotheses, the second is more parsimonious, because poorly documented historical phenomena need not be invoked. However, drier glacial climates in the east, a consequence of low winter rain and little or no summer rain ( Cowling ., 1999 ), would mean that conditions there would have been even more unfavourable than now for nonsprouters throughout most of the Pleistocene. Both of these hypotheses are consistent with the higher incidence of rare species in the west, and higher levels of beta and gamma diversity there, associated with the turnover of these rare species — as well as locally common habitat specialist species ( Cowling, 1990 ) — along environmental and geographical gradients, respectively ( Cody, 1986 ; Cowling ., 1992 ). These rare species do not contribute to community patterns; hence, biological heterogeneity showed no geographical or topographic patterns. The weak topography pattern of diversity in the west arises from higher speciation rates and lower extinction rates in the topographically complex mountains, rather than from the influence of environmental heterogeneity on diversity. Appendix1 Species–area data for sites and associated broad habitat units (BHUs) ( Cowling & Heijnis, 2001 ) used in the analyses Broad habitat unit Area (km 2 ) No. spp. West‐lowland 0.15 157 Hagelkraal 0.15 157 Cape Flats 0.20 210 Boland 0.68 373 Cape Flats 1.37 229 Blackheath 2.70 379 Boland 10.40 585 De Hoop, Potberg, Agulhas 180.00 1179 Elim, Springfield, Hagelkraal, Agulhas 1609.25 1751 East‐lowland 0.67 150 Knysna 0.67 150 Aloes 1.41 217 St Francis 3.36 173 Goukamma 20.55 380 Suurbraak 27.86 446 Riversdale, Canca, Albertinia, Stilbaai, Gouritz, Blanco 2860.0 1580 West‐montane 0.27 364 Franschhoek 0.27 364 Kogelberg 1.58 533 Franschhoek 1.82 483 Klein River 6.02 697 Klein River 14.46 773 Franschhoek 45.30 1142 Cape Peninsula 77.50 1036 Kogelberg 240.0 1383 Cape Peninsula 471 2256 Cederberg 1259 1175 East montane 6.17 313 Outeniqua 6.17 313 Zuurberg 207.8 1100 Rooiberg 250.0 481 Klein Swartberg, Little Karoo 340.0 473 Southern Langeberg 1737.8 1203 Kouga, Baviaanskloof 1778.2 1122 Appendix 2 Plant community data for broad habitat units (BHU) (see Cowling & Heijnis, 2001 ) in the Cape Floristic Region Broad habitat unit Area sampled (km 2 ) No. relevés No. communities Data source West lowland Langebaan 250 157 24 Boucher (1987 ) Cape Flats 175 60 11 Boucher (1987 ) Hopefield 345 139 16 Boucher (1987 ) Blackheath 110 20 6 Boucher (1987 ) Springfield 0.15 42 3 Richards . (1995 ) Hagelkraal 0.15 33 2 Richards . (1995 ) Swartland 420 42 11 Boucher (1987 ) Boland 480 66 16 Boucher (1987 ) Overberg 49 23 3 Kemper (1997 ) East lowland 220 42 4 Cowling (1982 ) St Francis 220 42 4 Cowling (1982 ) Humansdorp 175 42 6 Cowling (1982 ) Kromme 375 27 3 Cowling (1982 ) West montane Cederberg 1260 197 26 Taylor (1996 ) Swartruggens 127 125 9 Lechmere‐Oertel (1998 ) Hawequas 9.7 105 10 Van Wilgen & Kruger (1985 ) Franschhoek 0.4 201 5 McDonald (1988 ) Cape Peninsula 2 48 3 Glyphis . (1978 ) 1.24 53 2 Joubert & Moll (1982 ) 1.4 38 2 Laidler . (1978 ) 4 78 18 McKenzie . (1977 ) 77 87 10 Privett (1998 ) Kogelberg 1.6 367 11 Kruger (1974 ) 240 250 29 Boucher (1978 ) East montane Southern Langeberg 1748 299 46 McDonald . (1996 ) Groot Swartberg 100 50 9 Bond (1981 ) Outeniqua 150 65 11 Bond (1981 ) Kouga 400 75 8 Euston‐Brown (1995 ) Baviaanskloof 200 38 6 Euston Brown (1995 ) Cockscomb 50 28 5 Cowling (1982 ) Acknowledgments This study was funded by the Global Environment Facility (through World Wide Fund–South Africa) and the University of Port Elizabeth. Mike Rosenzweig, Rob Whittaker and two anonymous referees made useful comments on an earlier draft.

Journal

Diversity and Distributions – Wiley

Published: May 1, 2002

There are no references for this article.