The inselberg flora of Atlantic Central Africa. I. Determinants of species assemblages

Share Embed

Descrição do Produto

Journal of Biogeography (J. Biogeogr.) (2005) 32, 685–696


The inselberg flora of Atlantic Central Africa. I. Determinants of species assemblages Ingrid Parmentier1*, Tariq Ste´vart1 and Olivier J. Hardy2


Laboratoire de Botanique Syste´matique et de Phytosociologie, CP 169 and 2Laboratoire d’Eco-Ethologie Evolutive, CP 160/12 320, Universite´ Libre de Bruxelles, Brussels, Belgium


Aims To identify the relative contributions of environmental determinism, dispersal limitation and historical factors in the spatial structure of the floristic data of inselbergs at the local and regional scales, and to test if the extent of species spatial aggregation is related to dispersal abilities. Location Rain forest inselbergs of Equatorial Guinea, northern Gabon and southern Cameroon (western central Africa). Methods We use phytosociological releve´s and herbarium collections obtained from 27 inselbergs using a stratified sampling scheme considering six plant formations. Data analysis focused on Rubiaceae, Orchidaceae, Melastomataceae, Poaceae, Commelinaceae, Acanthaceae, Begoniaceae and Pteridophytes. Data were investigated using ordination methods (detrended correspondence analysis, DCA; canonical correspondence analysis, CCA), Sørensen’s coefficient of similarity and spatial autocorrelation statistics. Comparisons were made at the local and regional scales using ordinations of life-form spectra and ordinations of species data. Results At the local scale, the forest-inselberg ecotone is the main gradient structuring the floristic data. At the regional scale, this is still the main gradient in the ordination of life-form spectra, but other factors become predominant in analyses of species assemblages. CCA identified three environmental variables explaining a significant part of the variation in floristic data. Spatial autocorrelation analyses showed that both the flora and the environmental factors are spatially autocorrelated: the similarity of species compositions within plant formations decreasing approximately linearly with the logarithm of the spatial distance. The extent of species distribution was correlated with their a priori dispersal abilities as assessed by their diaspore types. Main conclusions At a local scale, species composition is best explained by a continuous cline of edaphic conditions along the forest-inselberg ecotone, generating a wide array of ecological niches. At a regional scale, these ecological niches are occupied by different species depending on the available local species pool. These subregional species pools probably result from varying environmental conditions, dispersal limitation and the history of past vegetation changes due to climatic fluctuations.

*Correspondence: Ingrid Parmentier, Laboratoire de Botanique Syste´matique et de Phytosociologie, ULB, CP 169, 50 Avenue F.D. Roosevelt, 1050 Bruxelles, Belgium. E-mail: [email protected]

Keywords Dispersal, canonical ordination analysis, Central Africa, environmental determinism, inselberg, island biogeography, spatial autocorrelation, vegetation gradients.

The relative importance of stochastic, deterministic and historical processes in determining species assemblages is much

debated (Chesson & Case, 1986; Ricklefs, 1987; Borcard et al., 1992; Husband & Barret, 1996; Tilman et al., 1997; Condit et al., 2002; Whitfield, 2002; Tuomisto et al., 2003). A classical point of view holds that local species composition is mostly driven by

ª 2005 Blackwell Publishing Ltd




I. Parmentier et al. niche differentiation according to competitive exclusion principles, giving much importance to deterministic biotic and abiotic ecological interactions (Schoener, 1986). The spatial pattern of species assemblage should then closely match environmental heterogeneity. Other theories, such as the ‘theory of island biogeography’ (MacArthur & Wilson, 1967) and its recent extension, the ‘unified neutral theory of biogeography and biodiversity’ (Hubbell, 2001), put more emphasis on stochastic processes such as dispersal and demographic drift, which can lead to colonization and local extinction, respectively. The particular species composition of a site is then viewed largely as a matter of chance but several characteristics of species assemblages remain predictable, such as the species–area curve or the species abundance distribution. In these ‘neutral’ models, the spatial pattern of species assemblage results from dispersal limitation, which can cause spatial autocorrelation of species occurrence independently of environmental variables. At a coarse geographical scale, the history of vegetation changes due to past climatic fluctuations is also recognized by biogeographers as a major determinant of species assemblages (Maley, 1996; Robbrecht, 1996; Sosef, 1996). Therefore, the floristic composition, diversity and spatial structure of the vegetation result from multiple causes, with different factors interacting and often resulting in an overlaid effect in space. As will be shown, inselbergs are model ecosystems for investigating the relative roles of these factors at different spatial scales. Inselbergs (‘islands hills’ in German) are rocky outcrops standing out from encircling plains as more or less isolated hills or groups of hills (Thomas, 1994). Due to their shallow soils, inselberg environmental conditions vary from very dry, with high temperatures and high levels of evapotranspiration (Szarzynski, 2000), to very humid where water flows and accumulate (i.e. down the slopes and in rock depressions). Hence, inselbergs generate a range of microhabitats where ecological conditions contrast moderately to extremely from the surrounding landscape (the ‘matrix’). The degree of contrast depends on the position of the local habitat along the inselberg– matrix ecotone as well as on the regional climate and edaphic characteristics of the matrix (Burke, 2003). In tropical rain forest landscapes, inselbergs can be regarded as habitat islands (Whittaker, 1998) providing shelter for an azonal type of vegetation associated with azonal climatic conditions. These extreme habitats often harbour endemic species and are refuges for species adapted to drier climatic conditions (Villiers, 1981; de Granville, 1982; Reistma et al., 1992; Parmentier, 2001). Several characteristics make rain forest inselbergs particularly interesting for studying the determinants of the vegetation and its flora, specifically for assessing the relative importance of deterministic, stochastic and historical processes shaping plant species assemblages at different scales. (1) There is a steep ecological gradient along the forest-inselberg ecotone affecting floristic composition (deterministic processes on a local scale). (2) The fragmented, island-like spatial distribution of the inselberg biotope suggests that dispersal (colonization) and demographic stochasticity (local extinction) might also be important determinants of the floristic composition (stochastic processes). (3) Differences in 686

climatic and/or edaphic conditions among inselbergs from within the same region allow investigation of the impact of largescale ecological gradients (deterministic processes on a regional scale). (4) Inselbergs may constitute refugia for both xeric and hydrophilic species during major climatic fluctuations (historical processes). To investigate species–environment relationships, multivariate statistical approaches have been developed to characterize the correlation between species data and environmental data (e.g. canonical correspondence analysis, CCA; ter Braak, 1987). A recurrent problem with these approaches is that the spatial autocorrelation of species data and environmental data can lead to overestimates of the real effect of ecological factors on species assemblages (Lennon, 2000; Legendre et al., 2002). Hence, to interpret observed species–environment correlations properly, it is important to assess the spatial structure of both environmental and species data (using spatial autocorrelation techniques) (Borcard et al., 1992; Lennon, 2000; Burke, 2002; Tuomisto et al., 2003). Another useful approach to assess the impact of environmental factors on species assemblages is to look at species functional characteristics because the latter are likely to be more directly dependent upon ecological constraints (Smith et al., 1997; Burke, 2001). When the organization of species assemblages is governed by limited dispersal and demographic drift rather than environmental heterogeneity, the patterns of species distribution can be predicted by neutral community models (e.g. Chave & Leigh, 2002). In this case, the spatial autocorrelation of a species occurrence is expected to decay according to the logarithm of the spatial distance (Chave & Leigh, 2002; Condit et al., 2002). The rate of decay, quantified by the slope of the autocorrelogram of species occurrence, which also expresses the extent of species spatial aggregation, is steeper when dispersal is more limited (Hardy & Sonke´, 2004). Hence, spatial autocorrelation approaches can be useful in comparing the patterns of spatial distribution between species, and to test if differences can be explained by species dispersal abilities. This is a first extensive study of the floristic composition and patterns of diversity of rain forest inselbergs from the western part of Central Africa. Our objective is to characterize the spatial organization of the inselberg flora at different scales of observation and to assess which factors might best explain the observed patterns. We will consider the impact of dispersal limitation, ecological factors and historical factors. This paper focuses on the spatial distribution of plant species and the patterns of differentiation of floristic assemblages according to ecological variables. Plant life-forms will be used as surrogates to describe functional diversity. We address the following questions: (1) Within an inselberg, do the main gradients in species and plant life-form compositions coincide with the ecological gradient along the forestinselberg ecotone? (2) Are the main gradients in species and plant life-form compositions correlated at the local and regional scale? (3) Do plant species distributions across inselbergs fit the patterns of regional ecological gradients? (4) Are plant species distributions affected by dispersal limitation? Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Spatial patterns of inselberg flora MATERIALS AND METHODS Study sites and inselberg plant formations Twenty-seven inselbergs were prospected in continental Equatorial Guinea (Rio Muni), northern Gabon and southern Cameroon (Fig. 1, Table 1). The study area extends c. 270 km in longitude and 280 km in latitude. Detailed climatic data are incomplete for the studied area. The general pattern is an equatorial climate with two dry and two wet seasons. Annual rainfall is affected by the distance to the ocean and the altitude, ranging from 1600 mm (BM, Cameroon) to more than 2500 mm (MA, EN, Equatorial Guinea). There are generally two dry months: July and August in the southern range of the study area, January and July in the central range, and December and January in the north-east. Temperatures remain between 23 and 26 C all year-round. These inselbergs are nowadays surrounded by primary rain forests, secondary forests or fields. Their altitude varies between 580 and 800 m and their elevation above the surrounding plains does not exceed 150 m. The rock outcrops of Monte Ale´n (MA) and Engong (EN) are exceptions as they are located in the Niefang mountain range, at 1130 m altitude. The choice of the inselbergs surveyed was strongly constrained by their accessibility. The studied inselbergs represent just a subset of the inselbergs of the investigated region. In Equatorial Guinea, almost all the inselbergs known by local people were investigated. In Cameroon and Gabon, only a few sites were investigated. As we are not aware of any study having recorded all the existing inselbergs, it is difficult to assess the proportion covered by our sampling outside Equatorial Guinea. We characterized the size of each inselberg by four ordinal classes based on its unforested area, which varies between < 1 and > 40 ha (Table 1). Following preliminary results obtained in three of these inselbergs (Parmentier et al., 2001), we considered six main plant formations along the forest-inselberg ecotone, recognized by their structural characteristics (independently of their floristic content). They typically form the following sequence from the surrounding forest to the bare rock: (1) a saxicolous

Figure 1 Location of the 27 inselbergs. Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Table 1 Characteristics of the 27 studied inselberg sites. Unforested area class (A) – 1: < 1 ha, 2: 1–10 ha, 3: 10–40 ha, 4: > 40 ha; number of phytosociological releve´s (NR); mean annual rainfall (AP); altitude (Al); longitude (N); latitude (E) Site



AP (mm)

Al (m)




3 3 3 2 3 3 3 3 4 4 3 1 4 2 2 1 3 3 4 3 2 2 1 4 4 2 2

23 20 22 4 12 7 31 40 39 18 32 3 31 19 17 4 9 17 7 12 17 10 9 58 15 3 3

1843 1950 2100 2000 1750 1750 2000 2500 1615 2000 2500 2000 1843 2500 2500 1900 2100 2100 2100 1843 2300 2000 1900 2000 1615 1615 2000

710 590 620 720 810 810 640 775 730 800 1120 610 740 1130 1130 670 650 780 680 725 760 800 830 745 660 615 600

0138.63¢ 0104¢ 0150.24¢ 0127.28¢ 0242.98¢ 0243.02¢ 0127.16¢ 0135.04¢ 0310.72¢ 0122.16¢ 0137.46¢ 0100.15¢ 0149.65¢ 0139.61¢ 0139.76¢ 0056.79¢ 152¢ 0150.96¢ 0151.62¢ 0150¢ 0119.07¢ 0127.88¢ 0100.1¢ 0127.45¢ 0318.02¢ 0316.97¢ 0100¢

1137.31¢ 1112¢ 1055.72¢ 1102.42¢ 1116.33¢ 1115.96¢ 1119.80¢ 1028.04¢ 1248.21¢ 1119.34¢ 1017.94¢ 1054.30¢ 1137.52¢ 1017.48¢ 1017.33¢ 1117.08¢ 1059¢ 1058.88¢ 1058.20¢ 1138¢ 1048.94¢ 1101.36¢ 1112.6¢ 1102.09¢ 1247.32¢ 1247.97¢ 1054¢

forest (sf) fringes the inselberg. Trees are generally smaller (< 22 m) and the structure is more open than in the surrounding forest on deeper soils. (2) The forest fringe (ff) has a width of c. 5 m and is the part of the forest where lateral sunlight still reaches the ground. It is composed of small trees (< 12 m), shrubs, herbaceous species and epiphytes. The abundance of epiphytes is due to the light availability and the accumulation of morning mist. (3) The shrubby fringe (sh) is mainly composed of 5 m high shrubs and a dense cover of lianas and herbaceous species. (4) The herbaceous fringe (hf) is c. 1.4 m high. (5) The humid grasslands (hg), c. 0.6 m high, are limited to sweeping zones and small depressions where soil depth is more important and humidity may be conserved between rainy episodes. (6) The dry grasslands (dg), which are c. 0.5 m high, grow on steeper slopes and on bare rock. Note that ‘grassland’ denotes here low vegetation dominated by herbaceous species (not necessarily Poaceae). A frequent major component of these grasslands in West and Central Africa is constituted by mats of the Cyperaceae Afrotrilepis pilosa (Bo¨ck.) J. Raynal. The same sequence of plant formations occurs at the summit of the inselberg when it is covered by a saxicolous forest. These saxicolous forests act as water reserves. The water sweeps gently from them almost continuously during the rainy season, allowing hygrophilous species to 687

I. Parmentier et al. develop on the slopes of the inselberg and in the grasslands fringing the upper saxicolous forest. Although defining these plant formations is somewhat arbitrary, as there is generally a continuous transition between them, they were used as surrogates to identify major habitat types.

(provided with fleshy outlier), sclerochores (lack of distinctive character but light), ballochores (forcibly ejected from parent plant), pogonochores (with plumose appendages, hairs, aigrettes), desmochores (hooked, spiny, bristeled) and pterochores (with scarious wing-like appendages).

Floristic data

Data analyses

Field data were collected from January 1999 to July 2002. This paper focuses on seven angiosperm families (Rubiaceae, Melastomataceae, Orchidaceae, Begoniaceae, Poaceae, Commelinaceae and Acanthaceae) and the Pteridophytes. These taxa constitute a representative sample of the inselberg flora as they comprise 36% of the species encountered in the whole flora of three of the sites (PN, AN and NB; Parmentier et al., 2001). Although our sampling was not limited to these taxa, we restricted our data analysis to them because their taxonomical identification was most reliable. The nomenclatural authority we followed is Lebrun & Stork (1995/1997, 2003) for the angiosperms and Kramer & Green (1990) for the pteridophytes. Data were gathered from 482 phytosociological plots (releve´s) investigated by I. Parmentier and a collection of c. 1450 herbarium samples and living plants (orchids) collected by I. Parmentier and T. Ste´vart, which are not necessarily part of the releve´s. In each inselberg, releve´s were located randomly within each of the plant formations present and accessible. This stratified sampling was chosen to account for the important floristic variation along the forest-inselberg grassland ecotone, and to allow comparison between releve´s from similar habitats. As the shape, extent and accessibility of each plant formation varied substantially among inselbergs, it was not possible to adopt a standard sampling area ensuring a wellrepresentative floristic list. Therefore, for each releve´ the investigated area was increased until almost no new species could be added after c. 15 min of search, or until the entire continuous patch occupied by the plant formation was investigated. However, most releve´s in the saxicolous forest are incomplete as no stabilization of the species number could be reached. The mean area surveyed per releve´ was 31 m2 on grassland formations, 55 m2 on herbaceous and shrubby fringes, 136 m2 on forest fringes and 371 m2 on saxicolous forests. The shape of the sampling plots tended to be linear for fringe-type formations, and rounded for the other formations. All releve´s being done by the same person, they are assumed to be homogeneous in terms of sampling effort. Species abundances were scored in the releve´s, but in this paper, only presence/absence of species was considered. The flora of the studied taxa totalled 490 species, 364 of them being recorded at least twice throughout the releve´s and herbariums. More than half of the species belong to the Orchidaceae (139), Rubiaceae (128) and pteridophytes (70). At least one specimen of each species was deposited in the herbarium of the Universite´ Libre de Bruxelles (BRLU). For each species, the type of diaspore was assessed following the definitions of Dansereau & Lems (1957). Seven diaspore types were represented: sporochores (very minute), sarcochores 688

Life-form spectra per plant formation The life-form of each species was assessed (phanerophyte, chamaephyte, hemicryptophyte, geophyte, therophyte, epiphyte or liana) to compute the percentages of species of each life-form per releve´ and the average percentages per plant formation.

Ordination of releve´s at local and regional scales To characterize the organization of species assemblages at a local scale, we ordinated the releve´s of each of five geographical groups of inselbergs ({MA1, MA2, EN}, {AK, MF1, MF2, MF3}, {PN, NB, AN, DU, AS}, {KU, MI, AB}, {BM, PO, RM}) using a detrended correspondence analysis (DCA) (Jongman et al., 1995). Each group contained inselbergs < 35 km apart, subjected to very similar climatic and edaphic conditions, so that species assemblages within group should not be affected by ecological gradients among inselbergs or historical factors related to past climatic changes. We then assessed how much of the variance of species occurrence data was correlated to the position of a releve´ along the forest-inselberg ecotone. To this end, plant formations were described by an ordinal variable ranging from 1 to 6 following the sequence described above and introduced as a passive environmental variable in the DCA (i.e. without constraining the ordination). The proportion of the species–environment relationship explained by the two first DCA axes was then examined. To compare the results obtained at local and regional scales, the same analysis was also performed after pooling all the groups of inselbergs. The same set of analyses was repeated using the life-form spectra per releve´ (i.e. the percentages of species belonging to each lifeform) rather than the species data. All analyses were carried out using canoco 4.0 (ter Braak & Smilauer, 2002). The floristic similarity between plant formations was evaluated by the Sørensen similarity coefficient (Legendre & Legendre, 2003) using the total species diversity in all sites.

The influence of environmental and spatial variables at a regional scale Canonical correspondence analysis and partial CCA were used to characterize species–environment relationships, taking spatial variables into account. The species data were constituted by the floristic lists of each inselberg on the basis of the releve´s and herbaria data. The environmental variables included quantitative variables (altitude, mean annual rainfall, grassland area, A. pilosa grassland area and the general slope of the inselberg) and binary nominal variables (buffalo grazing, traces of fire and Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Spatial patterns of inselberg flora presence of an upper saxicolous forest acting like a water reserve) defined per inselberg. The spatial variables were constituted by all terms of the inselberg geographical coordinates in a metric projection (Transverse Mercator) for a cubic trend surface regression (Legendre, 1990). Partial Monte Carlo permutation tests were used to assess the statistical significance of each environmental and/or spatial variable before introducing them in the ordination model (forward selection, ter Braak & Smilauer, 2002). The matrices were subjected to four separate CCA using: (1) species and environmental data; (2) species and spatial data; (3) species and environmental data with spatial data as covariables; and (4) species and spatial data with environmental data as covariables. The two last analyses are thus partial CCA and were used to partition the variance of the species data into components explained by environmental variables, spatial variables and their shared effect (Borcard et al., 1992). To this end we used the same number of statistically significant spatial and environmental variables to ensure the same degree of freedom for both classes of variables. Monte Carlo permutation tests (999 permutations) were performed to assess the significance of the eigenvalue of the first canonical axis and of the trace statistics (ter Braak & Smilauer, 2002).

Spatial autocorrelation of species occurrence The spatial distribution of each species was assessed by computing the spatial autocorrelation of species occurrence data in the 473 releve´s, considering only species found in at least two releve´s (221 species). For a pair of releve´s i, j, and a species s, the spatial autocorrelation coefficient of species occurrence is defined as a ratio of the covariance between the two sites over the variance observed among all study sites (as a Pearson correlation coefficient): rijs ¼ Covðxis ; xjs Þ=Varðxs Þ


where xsi ¼ 1 if species s occurs in releve´ i, otherwise xsi ¼ 0. This is equivalent to the Moran’s I coefficient (Cliff & Ord, 1981) except that pairwise coefficients are not averaged over predefined distance intervals at this stage. As we considered only presence/absence data, Varðxs Þ ¼ xs ð1  xs Þ, where xs is the mean xsi over all releve´s (i.e. all 473 releve´s), equation (1) can be written as: rijs ¼ ðxis  xs Þðxjs  xs Þ=ðxs ð1  xs ÞÞ


Averages over species were obtained by summing covariance terms and variance terms and then taking their ratio: rij ¼

X s

ðxis  xs Þðxjs  xs Þ


xs ð1  xs Þ



Actually, rij is also a relative measure of floristic similarity between releve´s i and j (rij is relative to the mean similarity between two sampled releve´s, which is zero) where the weight of a species is proportional to the variance of its occurrence. The rij values were averaged according to mutually exclusive distance classes, k, between releve´s: Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

IðkÞ ¼


, rij



dij 2k

where dij is the Euclidian distance separating i and j for releve´s from different sites (dij ¼ 0 for releve´s from the same site), and N(k) is the number of i, j pairs belonging to distance class k. Distance classes were defined as follows (in km): [0], ]0–10], ]10–20], ]20–50], ]50–100], ]100–200], ]200–500]. I(k) are actually weighted average Moran’s I coefficients (Cliff & Ord, 1981) and the plot of I(k) values against distance is an autocorrelogram which is expected to display a flat curve if species are randomly distributed, and a decreasing curve if species tend to be geographically aggregated. Hence, I(k) values provide a global picture of how floristic similarity between releve´s varies within (for the first distance interval) and among sites (other distance intervals). To compare the spatial structure of the flora from grassland habitats and from forested habitats, I(k) values were computed considering, in the first case, all releve´s taken in grasslands or herbaceous fringe, and in the second case, all releve´s taken in saxicoulous forest, forest fringe or shrubby fringe. The rsij or rij values were also regressed on ln(dij), giving the regression slope bs or b (pairs of releve´s within site were not considered when computing bs or b). The quantity bs synthesizes the extent of spatial aggregation of species s: the more negative it is, the stronger the spatial aggregation (Hardy & Sonke´, 2004). On the basis of all releve´s, we checked whether bs values varied among plant families or among diaspore types. Computations were done with the software SPAGeDi (Hardy & Vekemans, 2002), which was initially designed for genetic data analysis, by considering species as ‘loci’ and releve´s as ‘individuals’ (see Hardy & Sonke´, 2004).

Spatial autocorrelation of environmental data The spatial distribution of each environmental variable was assessed by the spatial autocorrelation of their value in the 27 sites. The method used is the same as for the spatial analysis of species distributions (equation 1), replacing xsi by the value of the environmental variable in site i. Nominal variables were transformed into binary form by the addition of dummy variables. RESULTS Life-form spectra The relative contribution of each life-form to the species richness per releve´ for each plant formation is depicted in Fig. 2. Naturally, phanerophytes are more represented at the forest end of the ecotone whereas hemicryptophytes, geophytes and therophytes occupy the grassland end. Chamaephytes are well represented in all plant formations and epiphytes increase in relative importance from grassland to more forested formations. It should be noted that species richness decreases substantially from the forest to the grassland formations 689

I. Parmentier et al. 100%

83.8% (range: 47.4–98.5%) of the relationship between species data and the sequence of plant formations (i.e. when plant formations are described by an ordinal variable representing the forest-inselberg ecotone; Table 2). Hence, plant formations consistently capture the major floristic gradient found within inselbergs. The ordination of releve´s always revealed some overlap between plant formations, especially between those adjacent along the forest-inselberg ecotone (Fig. 3), suggesting a continuous transition rather than discontinuous entities. This continuous transition is also suggested by the matrix of floristic similarity between plant formations when comparing the total list of species found in each plant formation (all inselbergs combined): the Sorensen’s coefficients are above 0.5 between adjacent plant formations in the forest-inselberg ecotone, and decrease regularly at values below 0.5 between plant formations further apart (Table 3). When the DCA was applied on all groups of inselbergs combined, the first axis expressed only 2.7% of the variance of species data and only 41.2% of the relationship between species data and plant formations (Table 2). Hence, at a larger scale (c. 200 km), other factors than the forest-inselberg ecological gradient become predominant in explaining species assemblages. When the DCA was applied on life-form spectra per releve´ rather than species data, at a local scale (i.e. within each group of nearby inselbergs), the first axis expressed 33.1% (range: 27.7–39.6%) of the variance in life-form spectra and 80.3% (range: 53.4–97.8%) of the relationship between life-form spectra and plant formations. After pooling the five groups of inselbergs, the first DCA axis still expressed 28.1% of the variance of life-form spectra and 93.8% of the relationship between life-form spectra and plant formations (Table 2).





G 40%











Figure 2 Relative contributions of each life-form to the species richness per releve´ according to plant formations: phanerophytes (P), chamaephytes (C), he´micryptophytes (H), geophytes (G), therophytes (T), epiphytes (E), other (O, including lianas and life-forms representing < 5% of the species of the plant formation).

(Table 3) so that, although chamaephytes constitute a gradually increasing proportion of the species richness along this gradient, the absolute species richness of chamaephytes decreases along this gradient. Ordinations of species and life-form data at local and regional scales Within each group of nearby inselbergs, the first axis of the DCA of the species data displayed the releve´s in a sequence matching the sequence of plant formation recognized (Fig. 3). Although only 7.7% (range: 6.5–10.1%) of the variance of the species data was expressed by the first axis, the latter expressed

Table 2 Main parameters of the DCA of species and life-form data per releve´ within five groups of nearby inselbergs and for all groups combined. The position of the releve´ within the sequence of plant formations recognized was introduced as a passive ordinal ‘environmental’ variable. Only main results for the two first axes are presented. Inselberg groups: G1 (AK, MF1, MF3), G2 (BM, PO, RM), G3 (KU, MI, AB), G4 (MA1, MA2, EN), G5 (PN, AS, NB, DU, AN) Ordination of species data








All sites

Species–environment correlation

1 2 1 2 1 2

0.937 0.039 10.1 4.2 96.7 0.01 7.387 8.9

0.750 0.619 7.4 5.8 47.4 30.1 12.272 5.9

0.860 0.076 7.7 5.1 98.5 0.0 9.090 6.3

0.783 0.127 6.5 4.0 87.1 0.9 11.238 4.4

0.806 0.100 6.6 4.3 89.1 0.2 10.890 4.6

0.827 0.192 7.7 4.7 83.8 25.6 10.175 6.0

0.529 0.311 2.7 2.2 41.2 7.4 29.875 1.9

1 2 1 2 1 2

0.794 0.007 32.8 14.6 96.3 0.0 1.634 21.1

0.444 0.234 27.7 11.3 68.0 1.2 2.715 7.7

0.829 0.095 39.6 10.6 97.8 0.3 1.461 26.4

0.397 0.397 31.1 15.0 53.4 33.1 1.852 10.3

0.787 0.24 34.2 13.4 85.9 2.9 1.583 23.3

0.650 0.195 33.1 13.0 80.3 7.5 1.849 17.8

0.753 0.141 28.1 12.9 93.8 0.0 2.03 16.7

Variance of species data (%) Variance of species–environment relationship (%) Sum of all eigenvalues Total variance explained by the environmental variable (%) Ordination of biological types Life-form/environment correlations Variance of life-form data (%) Variance of life-form/environment relationship (%) Sum of all eigenvalues Total variance explained by the environmental variable (%)


Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Spatial patterns of inselberg flora


Axis 2

Figure 3 DCA ordination diagram (axes 1 and 2) of 55 releve´s (72 species) of a group of nearby inselbergs (AK, MF1, MF2, MF3) showing that the first axis coincides with the gradient of plant formations (centroid positions of the latter are shown by crosses). Similar patterns were obtained for the other groups of inselbergs. sf: saxicolous forest; ff: forest fringe; sh: shrubby fringe; hf: herbaceous fringe; hg: humid grassland; dg: dry grassland.

sf ff


sh hf

sf hf


hg dg Axis 1

Table 3 Total number of species in each plant formation (bold); number of species in common between plant formations (below diagonal) and Sørensen’s similarity coefficient (above diagonal); sf: saxicolous forest; ff: forest fringe; sh: shrubby fringe; hf: herbaceous fringe; hg: humid grassland; dg: dry grassland; N: mean species number in the releve´s of each plant formation

sf ff sh hf hg dg N








184 124 91 54 19 16 7.3

0.67 286 172 102 46 34 12.7

0.43 0.65 244 115 57 41 10.1

0.32 0.46 0.58 154 61 43 7.3

0.15 0.26 0.36 0.54 71 36 4

0.14 0.20 0.28 0.42 0.59 52 2.6

The difference between the ordinations of species and lifeforms indicate that whereas plant formations capture well the gradients of both species data and life-form spectra at a local scale, over a large spatial extent they efficiently predict only life-form spectra. Influence of environmental and spatial variables at a regional scale When applying a CCA to the species data (species list per inselberg) using environmental variables only, low variance inflation factors (< 5) were obtained, indicating that each environmental variable contributed independently to the overall ordination. Both the eigenvalue of the first canonical axis and the trace statistic were significant (P ¼ 0.001). All variables together explained 41.8% of the total variability. After forward selection, three variables (altitude, rainfall and A. pilosa grassland area) were significantly related to the vegetation composition. These variables explained 22.8% of the total variance of species data (Table 4). This ordination is strongly influenced by two outsider groups of inselbergs (Fig. 4): the nearby inselbergs MA1, MA2 and EN (which are the most western) are distinguished mainly by their higher mean annual rainfall and altitude, and the nearby inselbergs PO, RM and BM (which are the most northern and eastern) by lesser rainfall. Part of the explained variation is likely due to the spatial structure of the environmental variables. When applying a Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Table 4 CCA of species data per inselberg using eight environmental variables Environmental variables

Marginal effect

Conditional effect

Altitude Rainfall A. pilosa grassland area Buffalo grazing General slope of the inselberg Upper saxicolous forest Grassland area Traces of fire

0.34 0.34 0.26 0.25 0.22 0.18 0.16 0.13

0.34 0.19 0.22 0.11 0.12 0.14 0.14 0.12

Conditional effects of significant environmental variables are in bold (P £ 0.05). % Variance explained by all variables: 41.8; % variance explained by significant variables: 22.8.

CCA on the nine spatial variables of a cubic trend surface regression (x, x2, x3, y, y2, y3, xy, x2y, xy2, where x and y are, respectively, the longitude and latitude of the sites expressed in UTM coordinates), forward selection kept x, x2, y, y2, xy2, x2y and y3 as significant variables, explaining 43.7% of the total variance of species data (Table 5). When these spatial variables were used as covariables to test the nine environmental variables, only the general slope of the inselberg and the A. pilosa grassland area remained significant after a forward selection, explaining 9% of the species variance. On the basis of partial CCA, variance partitioning between the three significant environmental variables (altitude, rainfall and A. pilosa grassland area) and the three best significant spatial variables (x, x2, y) yielded 15.5% of total variance explained by spatial variables, 13.4% by environmental variables and 9.4% of common effect. Spatial autocorrelation of species occurrence and of environmental variables The floristic similarity between releve´s decreased roughly linearly with the logarithm of the distance separating them (Fig. 5a). The observed patterns were similar for releve´s from grassland formations and for releve´s from forested formations. Releve´s within inselbergs were somewhat more similar than releve´s from inselbergs separated by < 10 or 20 km (Fig. 5a). As expected, releve´s taken in different types of plant 691

I. Parmentier et al. PO


RM BM I(k)













0.8 0.6



Axis 2













0.4 0.2 0.0 –0.2 1




Distance (km)


Axis 1 Figure 4 CCA ordination diagram (axes 1 and 2) for 27 inselbergs (298 species) and the following environmental variables: altitude (al), mean annual rainfall (pr), unforested area (ua), Afrotrilepis pilosa grassland area (af), general slope of the inselberg (sl), buffalo grazing (bu), traces of fire (fi), and presence of upper saxicolous forest acting like a water reserve (sf). The two axes explain 19% of total variability and 46% of the species–environment relationship. Table 5 CCA of species data per inselberg using nine spatial variables (x is longitude, y is latitude) Spatial variables

Conditional effect

x y xy x2 y2 xy2 x2 y x3 y3

0.34 0.19 0.12 0.29 0.18 0.17 0.18 0.13 0.14

Conditional effects of significant variables are in bold. % Variance explained by all variables: 52.8; % variance explained by significant variables: 43.7.

formations (grassland vs. forested) were much less similar than within type, but it is worth noting that the floristic similarity within a type of plant formations for inselbergs separated by more than 100 km reaches the level obtained between types (Fig. 5a). Hence, large-scale floristic differentiation within plant formations is as strong as small-scale differentiation between grassland formations and forested formations. 692

Figure 5 Spatial autocorrelation of (a) floristic data (i.e. floristic similarity between releve´s) and (b) environmental variables. For floristic data, we distinguished releve´s within grassland plant formations (triangles), within forested formations (circles), and between grassland and forested formations (crosses). The symbols on the vertical axis indicate the mean floristic similarity between releve´s from the same inselberg. For environmental variables, we report the weighted means of the autocorrelation values obtained for each variable, where each variable was weighted by its marginal effect in the CCA.

Five environmental variables were spatially autocorrelated, the slopes of their correlograms (b values) being significantly negative (P < 0.05): altitude, mean annual rainfall, buffalo grazing, A. pilosa grassland area, general slope of the inselberg. Note that the average autocorrelogram of the environmental variables, where each variable was weighted by its marginal effect in the CCA (Fig. 5b), is similar in shape to the autocorrelograms of the floristic data (Fig. 5a). The extent of species spatial aggregation, bs, varied significantly among diaspore types (Kruskal–Wallis test, P ¼ 0.025), the sclerochorous species displaying more aggregation than the sarcochorous and sporochorous ones (Fig. 6). This makes sense as the sarcochorous and sporochorous species are likely to be better dispersed, by animals and wind, respectively. It also varied among taxa (Kruskal–Wallis test, P ¼ 0.036), the Melastomataceae and Poaceae showing the strongest aggregation, and the Orchidaceae the weakest. As diaspore type is a rather conserved trait phylogenetically, we tried to distinguish the impact of diaspore type and family using a two-way factorial anova, after normalization of bs by the following transformation: log()bs + 0.005). To get data for each combination of family and diaspore type, we had to restrict the analyses to the Commelinaceae, Melastomataceae and Rubiaceae, where sarcochorous and sclerochorous species were represented. The diaspore type effect was significant (F1,55 ¼ 4.448, P ¼ 0.0395; mean bs ¼ )0.0096 for sarcochores, )0.0222 for sclerochores), Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Spatial patterns of inselberg flora –0.025


–0.020 –0.015 –0.010 –0.005 (9 2) re

(5 7) or oc ho Sp

ch o


re Sa rc o

ch o al lo B

Sc le ro ch o


(5 3)

(9 )


Figure 6 Comparison of the extent of species spatial aggregation according to their diaspore types. The spatial aggregation was quantified by the slope of correlograms, bs, and the mean values per diaspore type are shown. Error bars indicate approximate confidence intervals using twice the standard error. In parentheses are the numbers of species per diaspore type (diaspore types represented by less than five species were not considered).

whereas the family effect (F2,55 ¼ 0.5015, P ¼ 0.6083), and the interaction (F2,55 ¼ 0.7782, P ¼ 0.4643) were not. DISCUSSION Determinants of plant communities at a local scale Plant life-forms can be taken as surrogates for functional diversity (Smith et al., 1997) so that it is interesting to compare the ordination of releve´s based on their life-form spectra with the ordination of the same releve´s based on their floristic list. In both cases, the first axis expressed the steep ecological gradient of the forest-inselberg ecotone, even if it expressed a much higher proportion of the variance of life-form spectra than of species data (33.1% vs. 7.7%). The latter result is partially explicable by the difference in number of variables (eight life-forms compared with 50–80 species) but it also indicates that there is much stochasticity with respect to the presence/absence of a species in a releve´. Nevertheless, the fact that both species diversity and functional diversity are closely correlated with habitats (plant formations) strongly suggests that niche differentiation is a major determinant of species assemblages at this local scale. In three of the studied inselbergs (PN, NB, AN), Parmentier (2003) showed that the ecological gradient of the forest-inselberg ecotone is characterized by a decreasing soil depth and an increasing pH gradient from the saxicolous forest to the grasslands. Soil depth has also been identified as the main ecological gradient in rain forest inselbergs in Gabon (Reistma et al., 1992) and French Guiana (Sarthou, 2001). Determinants of plant communities at a regional scale When the DCA was applied on the pooled set of releve´s of all five groups of inselbergs (regional scale), the first axis still captured well the forest-inselberg ecotone when ordinating the life-form spectra, but not so well when ordinating the species Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

data, the latter being much influenced by the geographical position of the inselbergs. Hence, whereas the position along the forest-inselberg ecotone is the main variable explaining species assemblages at a local scale, other factors must be invoked at the regional scale, such as large-scale ecological gradients, the history of vegetation changes and limited plant dispersal abilities. Our results are similar to those found by Burke (2001) for inselbergs from the Namib desert: over large spatial extents, floristic composition was largely influenced by geographical position whereas growth forms seemed to be more dependent upon environmental factors. The strong spatial structure we observed in the rain forest inselberg flora could possibly be due to the insular character of a dry inselberg vegetation embedded in a humid matrix. Indeed, in tropical rain forests, the inselberg flora varies substantially between sites or group of sites, both in species composition and species abundance (e.g. Brazil: Safford & Martinelli, 2000; Equatorial Guinea: Parmentier, 2001; French Guyana: Gasc et al., 1998; Sarthou & Villiers, 1998; Sarthou et al., 2003; Gabon: Reistma et al., 1992). In contrast, in areas where the matrix and the inselberg environment are less contrasted, the inselberg flora is relatively homogenous among sites, at least for its characteristic elements occurring within the open vegetation (Porembski, 2000; Oumorou & Lejoly, 2003). Within habitat types, at a regional scale, the floristic similarity between releve´s decreases roughly linearly with the logarithm of the spatial distance. The origin of the spatial autocorrelation of species occurrence could possibly be due to: (1) spatially autocorrelated environmental factors; (2) dispersal limitation: floristic similarity among plots decreases with increasing geographical distance because of spatially limited species dispersal (cf. neutral theory, Hubbell, 2001); (3) the shared history of nearby inselbergs with respect to past vegetation changes. These non-exclusive hypotheses are discussed in turn below. The CCA we applied to all sites and environmental data (without taking into account the effect of spatial data) revealed three ecological variables significantly correlated with species data: altitude, rainfall and A. pilosa grassland area. However, these variables were spatially autocorrelated and the ordination was strongly influenced by two outsider groups of inselbergs located at different extremities of the region investigated (MA1, MA2, EN on the western side and PO, RM, BM on the north-eastern side) and displaying extreme values for the altitude (high for MA1, MA2, EN) or the mean annual rainfall (low for PO, RM, BM). Therefore, the question is whether the significant correlations with environmental variables reflect causal relationships or are a by-product of the strong spatial autocorrelation of both species and environmental variables (Lennon, 2000). The shape of the spatial autocorrelograms of the environmental variables and the species data were quite similar (Fig. 5a,b), but this similarity does not demonstrate a causal relationship. The problem is that the geographical scale of variation of environmental variables and species data were similar, so that a match between their respective coarse-scale patterns could be caused by chance. This problem is often 693

I. Parmentier et al. encountered in experimental studies trying to decipher species–environment relationships from the respective patterns of distribution at a regional scale (Lennon, 2000; Legendre et al., 2002). In our case, spatial variables were as good as (or even better than) environmental variables in accounting for patterns of variation in a statistical sense (Tables 4 and 5). One way to circumvent this problem is by partialling out some of the spatial effects by introducing spatial covariables in the CCA. With this approach, only the slope and the surface of A. pilosa grassland remained significant explanatory variables. This makes sense because whole communities are dependent on A. pilosa mats and some species are growing almost only on these mats (Porembski et al., 1996; Porembski, 2000; Parmentier et al., 2001; Ste´vart et al., in press). These monocotyledonous mats store a considerable amount of rainwater which would otherwise be lost by runoff (Biedinger et al., 2000). During the rainy season, there is a continuous seepage of water out of the mats and several species might be dependent on that runoff of seepage water to resist drought between the rainy episodes. It must be noted that we may have missed important environmental variables in our analysis, such as the human impact or the abundance of mist, for which we lacked reliable information. The interpretation of the CCA and the partition of variance that attributed similar impact of environmental and spatial variables must therefore be taken with caution. The slope aspect was not taken into account because it showed no effect in a previous study of three of the sites (Parmentier, 2003). Our spatial autocorrelation analyses revealed a significant difference in the extent of species aggregation according to species dispersal abilities, as inferred from their type of diaspore. A significant difference also occurred among families, but when both family and diaspore type were taken into account in a factorial two-way anova (on a restricted data set), only the effect of the diaspore type remained significant. This effect might be attributed to the easier long-distance propagule dispersal of sporochores and sarcochores. However, the low spatial structure of sporochores, which are represented by orchids and pteridophytes, could also be due to the fact that many inselberg orchids and pteridophytes are also epiphytes in the rain forest canopy. In Atlantic Central Africa, Ste´vart (2003) noticed that 72% of the orchids of the inselberg dry grassland and 85% of those of the inselberg herbaceous fringes are facultative epiphytes. Consequently, the inselbergs are not necessarily an insular ecosystem for these species. In any case, our results suggest that dispersal limitation is a major factor shaping plant communities at a regional scale. The pattern of autocorrelation is also consistent with neutral theoretical models that predict an approximate linear decrease of floristic similarity with the logarithm of the spatial distance (Hubbell, 2001). Such a pattern has also been observed for rain forest tree species in Amazonia (Condit et al., 2002) and central Africa (Hardy & Sonke´, 2004). A lesser correlation of floristic distance with geographical distance for pteridophytes than for Melastomataceae was also observed in the Amazonian rain forest and this difference was attributed to the greater dispersal limitation of Melastomataceae species (Tuomisto et al., 2003). 694

The ecological contrast between the rain forest matrix and the inselberg habitats increases along the forest-inselberg ecotone. We might therefore expect that the insular nature of inselbergs is more pronounced for species occurring in grassland formations, so that the resulting dispersal limitation could cause higher floristic differentiation among inselbergs for grassland than forested formations. However, this hypothesis is not confirmed by the spatial autocorrelation analyses as autocorrelograms were very similar for forested formations and grassland formations (Fig. 5a). This may mean that dispersal limitation is not much dependent on the inselberg– matrix ecological contrast, or that there is some compensation effect, for example, if species from grassland formations tend to have better long-distance dispersal abilities. Another explanation could be that ecological niches of the grasslands are only accessible to inselberg specialists, explaining why grasslands species have broad distributions. In the forested plant formations, niches are accessible to many species and may be occupied by whatever the surrounding matrix provides, leading to a subregional differentiation. To investigate such questions further, species affinities between inselbergs and the surrounding matrix must be established, as well as a better characterization of the matrix itself. The floristic geographical variation might also be profoundly influenced by the history of the vegetation in the region. Inselbergs are very old landscape elements that have undergone many climatic changes. During the last Pleistocene glaciations (c. 75,000–12,000 years bp), our study area was probably covered mostly by savannahs and the forested area strongly reduced (Maley, 1996; Leal, 2001). The location of the inselbergs relative to the Pleistocene tropical forest refuges has probably an influence on their present-day vegetation. Here again, further investigations of the flora of the surrounding matrix will be needed to better describe the chorological context in which the inselberg flora developed. CONCLUSION At a local scale, the inselberg flora is strongly influenced by environmental constraints, which are closely related to the position in the ecotone between the forest and the inselberg grasslands (deterministic factors). At a regional scale, the corresponding ecological niches are occupied by different species depending on the available local species pool. These subregional species pools probably result from varying environmental conditions, dispersal limitation and the history of past vegetation changes due to climatic fluctuations. Our results indicate that dispersal limitation (a stochastic factor) seems to play a role in the spatial pattern of the floristic composition. We also observed a regional variation in the importance of environmental variables (large-scale deterministic factors). Nevertheless, the strong spatial autocorrelation of both species data and environmental data renders the interpretation of the observed species–environment correlations very delicate. Quantifying the historical effect on the present species distribution is very difficult as the history of the vegetation in central Africa is Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Spatial patterns of inselberg flora still largely unknown. These three factors (stochastic, deterministic and historical) result in a gradual turnover in species pools that should be considered in conservation projects of the central African rain forest inselberg vegetation. Although this study is based on a subset of the flora, we think that our main results should apply for the whole flora. Once the floristic determinations are completed, we expect nevertheless to gather new insights by increasing the power of statistical analyses. ACKNOWLEDGEMENTS This work has benefited from the help of several people and institution. We thank the national herbaria of Cameroon, Gabon, Equatorial Guinea, Belgium, the projects ECOFAC, CUREF, DIVEAC, the WCS, the WWF and the FNRS. We are very grateful to the following botanists who checked the determination of the herbarium material: E. Robbrecht and J. Degreef (Rubiaceae), R. Faden (Commelinaceae), M. S. M. Sosef and J. J. F. E. de Wilde (Begoniaceae), P. J. Cribb and D. Geerinck (Orchidaceae), D. Champluvier (Acanthaceae), L. Pauwels (Poaceae) and D. Geerinck (Melastomataceae), E. Figueiredo (Pteridophytes). We also thank Dr John T. Hunter as well as two anonymous referees for their useful comments. REFERENCES Biedinger, N., Porembski, S. & Barthlott, W. (2000) Vascular plants on inselbergs: vegetative and reproductive strategies. Inselbergs. Biotic diversity of isolated rock outcrops in tropical and temperate regions (ed. by S. Porembski and W. Barthlott), pp. 117–140. Springer-Verlag, Berlin. Borcard, D., Legendre, P. & Drapeau, P. (1992) Partialling out the spatial component of ecological variation. Ecology, 73, 1045–1055. ter Braak, C.F.J. (1987) The analysis of vegetation–environment relationships by canonical correspondence analysis. Vegetatio, 64, 159–160. ter Braak, C.J.F. & Smilauer, P. (2002) CANOCO reference manual and CanoDraw for windows user’s guide. Microcomputer Power, Ithaca, NY. Burke, A. (2001) Determinants of inselberg floras in arid Nama Karoo landscapes. Journal of Biogeography, 28, 1211–1220. Burke, A. (2002) Plant communities of a Central Namib Inselberg landscape. Journal of Vegetation Science, 13, 483–492. Burke, A. (2003) Inselbergs in a changing world – global trends. Diversity and Distributions, 9, 375–383. Chave, J. & Leigh, E.G. (2002) A spatially-explicit neutral model of beta-diversity in tropical forests. Theoretical Population Biology, 62, 153–168. Chesson, P.L. & Case, T.J. (1986) Overview: nonequilibrium community theories: chance, variability, history and coexistence. Community ecology (ed. by J. Diamond and T.J. Case), pp. 229–239. Harper & Row, New York. Cliff, A.D. & Ord, J.K. (1981) Spatial processes: models and applications. Pion, London. Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Condit, R., Pitman, N., Leight, E.G. Jr., Chave, J., Terborgh, J., Foster, R.B., Nu´n˜ez, P.V., Aguilar, S., Valencia, R., Villa, G., Muller-Laudau, H.C., Losos, E. & Hubbell, S.P. (2002) Betadiversity in tropical forest trees. Science, 295, 666–669. Dansereau, P. & Lems, K. (1957) The grading of dispersal types in plant communities. Contributions de l’Institut de Botanique de Montreal, 71, 1–52. Gasc, J.P., Sarthou, C., Garrouste, R., Villiers, J.F., Cremers, G. & Thiollay, J.M. (1998) Inselbergs et savanes-roches en Guyane: biodiversite´ et conservation des milieux associe´s aux affleurements granitiques. Journal d’Agriculture Traditionnelle Tropicale et Botanique Applique´e, 40, 311–327. de Granville, J.J. (1982) Rain forest and xeric flora refuges in French Guiana. Biological diversification in the tropics (ed. by G.T. Prance), pp. 159–181. Columbia University Press, New York. Hardy, O.J. & Sonke´, B. (2004) Spatial pattern analysis of tree species distribution in a tropical rain forest of Cameroon: assessing the role of limited dispersal and niche differentiation. Forest Ecology and Management, 197, 191–202. Hardy, O.J. & Vekemans, X. (2002) SPAGeDi: a versatile computer program to analyse spatial genetic structure at the individual or population levels. Molecular Ecology Notes, 2, 618–620. Hubbell, S.P. (2001) The unified neutral theory of biodiversity and biogeography. Princeton University Press, Princeton, NJ. Husband, B.C. & Barret, S.C.H. (1996) A metapopulation perspective in plant population biology. Journal of Ecology, 84, 461–469. Jongman, R.H.G., ter Braak, C.J.F. & Van Tongeren, O.F.R. (1995) Data analysis in community and landscape ecology. Cambridge University Press, Cambridge, MA. Kramer, K.U. & Green, P.S. (1990) Pteridophytes and gymnosperms. The families and genera of vascular plants (ed. by K. Kubitzki), pp. 1–404 Springer-Verlag, Berlin. Leal, M.E. (2001) Microrefugia, small ice age forest remnants. Systematics and Geography of Plants, 71, 1073–1077. Lebrun, J.P. & Stork, A.L. (1995/1997) Enume´ration des plantes a` fleur d’Afrique tropicale, Vols III and IV. Editions des Conservatoire et Jardin botaniques de la ville de Gene`ve, Geneva. Lebrun, J.P. & Stork, A.L. (2003) Tropical African flowering plant. Ecology and distribution, Vol. 1. Editions des Conservatoire et Jardin botaniques de la ville de Gene`ve, Geneva. Legendre, P. (1990) Quantitative methods and biogeographic analysis. Evolutionary biogeography of the marine algae of the North Atlantic (ed. by D.J. Garbary and R.R. South), pp. 9–34. Springer-Verlag, Berlin. Legendre, P. & Legendre, L. (2003) Numerical ecology, 2nd English edn. Elsevier Science, Amsterdam. Legendre, P., Dale, M.R.T., Fortin, M.-J., Gurevitch, J., Hohn, M. & Myers, D. (2002) The consequences of spatial structure for the design and analysis of ecological field surveys. Ecography, 25, 601–615. Lennon, J.J. (2000) Red-shifts and red herrings in geographical ecology. Ecography, 23, 101–113. 695

I. Parmentier et al. MacArthur, R.H. & Wilson, E.O. (1967) The theory of island biogeography. Princeton University Press, Princeton, NJ. Maley, J. (1996) The African rain forest – main characteristics of changes in vegetation and climate from the Upper Cretaceous to the Quaternary. Essays on the ecology of the Guinea-Congo rain forest (ed. by I.J. Alexander, M.D. Swaine and R. Watling), pp. 30–73. Royal Society of Edinburgh, Edinburgh. Oumorou, M. & Lejoly, J. (2003) Aperc¸u de la ve´ge´tation de quelques inselbergs du Be´nin. Systematics and Geography of Plants, 73, 215–236. Parmentier, I. (2001) Premie`res e´tudes sur la diversite´ ve´ge´tale des inselbergs de Guine´e Equatoriale continentale. Systematics and Geography of Plants, 71, 911–922. Parmentier, I. (2003) Study of the vegetation composition in three inselbergs from Continental Equatorial Guinea (western central Africa): effects of site, soil factors and position relative to the lower or upper forest fringe. Belgian Journal of Botany, 136, 63–72. Parmentier, I., Lejoly, J. & Nguema, N. (2001) La ve´ge´tation des inselbergs du monument naturel de Piedra Nzas (Guine´e Equatoriale continentale). Acta Botanica Gallica, 148, 341–365. Porembski, S. (2000) West African Inselberg vegetation. Inselbergs. Biotic diversity of isolated rock outcrops in tropical and temperate regions (ed. by S. Porembski and W. Barthlott), pp. 117–211. Springer-Verlag, Berlin. Porembski, S., Brown, G. & Barthlott, W. (1996) A species-poor tropical sedge community: Afrotrilepis pilosa mats on inselbergs in West Africa. Nordic Journal of Botany, 16, 239–246. Reistma, J.M., Louis, A.M. & Floret, J.J. (1992) Flore et ve´ge´tation des inselbergs et dalles rocheuses: premie`re e´tude au Gabon. Bulletin du Museum National d’Histoire Naturelle, 14, 73–97. Ricklefs, R.E. (1987) Community diversity: relative roles of local and regional processes. Science, 235, 167–171. Robbrecht, E. (1996) Geography of African Rubiaceae with reference to glacial rain forest refuges. The biodiversity of African plants (ed. by L.J.G. van der Maesen, X.M. van der Burgt and J.M. van Medenbach de Rooy), pp. 564–581. Kluwer Academic Publishers, The Netherlands. Safford, H.D. & Martinelli, G. (2000) Southeast Brazil. Inselbergs. Biotic diversity of isolated rock outcrops in tropical and temperate regions (ed. by S. Porembski and W. Barthlott), pp. 339–389. Springer-Verlag, Berlin. Sarthou, C. (2001) Plant communities on a granitic outcrop. Nouragues: dynamics and plant–animal interactions in a neotropical rain forest (ed. by F. Bongers, P. Charles-Dominique, P.M. Forget and M. The´ry), pp. 65–78. Kluwer, Dordrecht. Sarthou, C. & Villiers, J.F. (1998) Epilithic plant communities on inselbergs in French Guiana. Journal of Vegetation Science, 9, 847–860. Sarthou, C., Villiers, J.F. & Ponge, J.F. (2003) Shrub vegetation on tropical granitic inselbergs in French Guiana. Journal of Vegetation Science, 14, 645–652. Schoener, T.W. (1986) Resource partitioning. Community ecology: patterns and process (ed. by J. Kikkawa and D.J. 696

Anderson), pp. 91–126. Blackwell Scientific Publications, Melbourne. Smith, T.M., Shugart, H.H. & Woodward, F.I. (eds) (1997) Plant functional types: their relevance to ecosystem properties and global change. Cambridge University Press, New York. Sosef, M.S.M. (1996) Begonias and African rain forest refuges. The biodiversity of African plants (ed. by L.J.G. van der Maesen, X.M. van der Burgt and J.M. van Medenbach de Rooy), pp. 602–611. Kluwer Academic Publishers, The Netherlands. Ste´vart, T. (2003) Taxonomical, ecological and phytosociological study of the Orchidaceae in Atlantic Central Africa. PhD in Biological Science, Universite´ Libre de Bruxelles, Brussels. Ste´vart, T., Ngok Banak, L. & Sosef, M.S.M. (in press) Synthe`se des inventaires re´alise´s sur les Orchidaceae dans le cadre du Projet d’Evaluation des Aires Prote´ge´es du Gabon. Proceedings du XVIIe`me Congre`s de l’AETFAT. Szarzynski, J. (2000) Xeric islands: environmental conditions on inselbergs. Inselbergs. Biotic diversity of isolated rock outcrops in tropical and temperate regions (ed. by S. Porembski and W. Barthlott), pp. 37–48. Springer-Verlag, Berlin. Thomas, M.F. (1994) Geomorphology in the tropics. A study of weathering and denudation in low latitudes. John Wiley & Sons Ltd, UK. Tilman, D., Lehman, C.L. & Yin, C.J. (1997) Habitat destruction, dispersal and deterministic extinctions in competitive communities. American Naturalist, 149, 407–435. Tuomisto, H., Ruokoleinen, K. & Yli-Halla, M. (2003) Dispersal, environment, and floristic variation of western Amazonian forests. Science, 299, 241–244. Villiers, J.F. (1981) Formations climaciques et relictuelles d’un inselberg inclus dans la foreˆt dense camerounaise. PhD thesis in Science, Muse´um National d’Histoire Naturelle, Universite´ Pierre et Marie Curie, Paris. Whitfield, J. (2002) Neutrality versus the niche. Nature, 417, 480–481. Whittaker, R.J. (1998) Island biogeography: ecology, evolution, and conservation. Oxford University Press, Oxford.

BIOSKETCHES Dr Ingrid Parmentier is a research assistant, with a doctorate degree on the biogeography, phytosociology and ecology of the inselberg vegetation in western central Africa. Dr Tariq Ste´vart has studied tropical African orchids since 1997. His research interest include orchid systematics, ecology of epiphytes, and historical and island biogeography. Dr Olivier Hardy mainly works on the spatial patterns of genetic diversity within plant populations and of species diversity within plant communities. He is interested in linking the concepts and methods developed in population genetics and community ecology.

Editor: Philip Stott Journal of Biogeography 32, 685–696, ª 2005 Blackwell Publishing Ltd

Lihat lebih banyak...


Copyright © 2017 DADOSPDF Inc.