1 The cultivation and dissemination of alien ornamental plants increases their potential to invade. More specifically, species with bird-dispersed seeds can potentially infiltrate natural nucleation processes in savannas.
2To test (i) whether invasion depends on facilitation by host trees, (ii) whether propagule pressure determines invasion probability, and (iii) whether alien host plants are better facilitators of alien fleshy-fruited species than indigenous species, we mapped the distribution of alien fleshy-fruited species planted inside a military base, and compared this with the distribution of alien and native fleshy-fruited species established in the surrounding natural vegetation.
3Abundance and diversity of fleshy-fruited plant species was much greater beneath tree canopies than in open grassland and, although some native fleshy-fruited plants were found both beneath host trees and in the open, alien fleshy-fruited plants were found only beneath trees.
4Abundance of fleshy-fruited alien species in the natural savanna was positively correlated with the number of individuals of those species planted in the grounds of the military base, while the species richness of alien fleshy-fruited taxa decreased with distance from the military base, supporting the notion that propagule pressure is a fundamental driver of invasions.
5There were more fleshy-fruited species beneath native Acacia tortilis than beneath alien Prosopis sp. trees of the equivalent size. Although there were significant differences in native plant assemblages beneath these hosts, the proportion of alien to native fleshy-fruited species did not differ with host.
6Synthesis. Birds facilitate invasion of a semi-arid African savanna by alien fleshy-fruited plants, and this process does not require disturbance. Instead, propagule pressure and a few simple biological observations define the probability that a plant will invade, with alien species planted in gardens being a major source of propagules. Some invading species have the potential to transform this savanna by overtopping native trees, leading to ecosystem-level impacts. Likewise, the invasion of the open savanna by alien host trees (such as Prosopis sp.) may change the diversity, abundance and species composition of the fleshy-fruited understorey. These results illustrate the complex interplay between propagule pressure, facilitation, and a range of other factors in biological invasions.
Dispersal of seeds by birds is essential for the maintenance of tree and shrub populations and for structuring vegetation in many ecosystems (Kollmann 2000). By perching and nesting in trees, frugivorous birds direct seeds of fleshy-fruited plants to subcanopy sites (Snow 1976). Such dispersal may lead to changes in vegetation that, depending on the management perspective and the plant species involved, may be viewed as succession (Cowling et al. 1997), recolonization after anthropogenic disturbances (Milton et al. 1997; Holl 2002), or invasion (Gimeno & Vila 2002; Buckley et al. 2006).
Birds are particularly important for determining the clumped distribution of fleshy-fruited plant species (Bews 1917; Archer et al. 1988). Where tree cover is incomplete, establishment of fleshy-fruited plants occurs predominantly under tree canopies, and far less in the open, because frugivorous birds perch in tree canopies (Debussché & Isenmann 1994; Vierera et al. 1994; Pausas et al. 2006) and subcanopy microsite conditions are favourable for seed germination and seedling establishment (Belsky et al. 1993; Verdú & García-Fayos 1996). The process of shrub-clump formation resulting from the distribution of seed of a number of fleshy-fruited species to a single perch site is known as ‘nucleation’ (viz. Pausas et al. 2006), because such clumps tend to grow around a single founder plant (or nucleus).
In savanna systems worldwide, isolated trees represent focal points for bird activity (Dean et al. 1999), for recruitment of bird-dispersed plants (Smith & Goodman 1987; Flores-Flores & Yeaton 2000), and for the development of species-rich clumps of fleshy-fruited plants (Archer et al. 1988; Skarpe 1991). Typically, the interactions between herbivory, avian seed dispersal and fire lead to the cycle, shown in Fig. 1, whereby a wind- or mammal-dispersed tree establishes in the grassland, attracting birds, many of which feed on both insects and fruits. The birds transport seeds of a variety of fleshy-fruited plants to a perch site in the tree. Shade-adapted seedlings establish below the canopy, forming a bush clump that attracts frugivorous birds that bring more seeds. Over time, fleshy-fruited species may grow to overtop the original host tree, outcompeting it for light (Yeaton & Manzanares 1986; Flores-Flores & Yeaton 2000). Eventually senescence, with or without fire, returns the clump to grassland.
In theory, the addition of new (i.e. alien) bird-dispersed plant species could alter the processes shown in Fig. 1 in many ways. The timing and abundance of propagule availability could clearly mediate the type and magnitude of influence of the added species. For example, artificially elevated propagule pressure (due to the widespread planting, feeding and watering of ornamentals and copious seed production in the absence of natural enemies) could enhance the ability of alien species to both establish and to then influence prevailing processes (Lockwood et al. 2005). In a meta-analysis of characteristics of invasiveness, Colautti et al. (2006) found propagule pressure was consistently associated with invasion success. Both for this reason, and because propagule pressure models invasions as a probabilistic process, Colautti et al. (2006) suggested that it should form the basis of a null model for biological invasions.
In this paper the terms ‘alien’, ‘native’ and ‘invasive’ follow definitions proposed by Pyšek et al. (2004). Native plants are ‘taxa that have originated in a given area without human involvement or that have arrived there without intentional or unintentional intervention of humans from an area in which they are native’, whereas alien taxa are those ‘whose presence [in a given area] is due to intentional or unintentional human involvement, or which have arrived there without the help of people from an area in which they are alien’. Invasive plants are a subset of naturalized plants that produce and disperse many reproductive offspring, and thus have the potential to spread over a large area.
The relationship between birds and fleshy-fruited plants is so widespread and general that new mutualisms (Richardson et al. 2000a) rapidly develop between birds (native and alien) and alien fleshy-fruited species, enabling the birds to disperse the plant propagules over long distances from established gardens to suitable microsites in natural vegetation. The addition of new (alien) tree species to an ecosystem provides further, perhaps alternative, foci for the deposition and establishment of fleshy-fruited alien species. Such mutualisms are clearly crucial facilitators of invasions. Simberloff & Von Holle (1999) suggest that synergistic interactions among invaders that accelerate invasions and/or amplify their effects on native communities can be seen as ‘invasional meltdown’. The important effects that invasions may have on naturally occurring mutualisms are less widely appreciated, but have momentous implications for biodiversity in invaded ecosystems (Traveset & Richardson 2006).
Arid ecosystems in most regions have been less affected by biological invasions than more mesic ecosystems (Loope et al. 1988; Milton et al. 1999). This may be because the subset of drought-hardy plants is smaller. Other contributing reasons may be that establishment opportunities (for example the occurrence of substantial postgermination rains) are more rare (and therefore slower to become evident) and/or that fewer species and individuals of alien species have been introduced to arid, as opposed to more mesic, regions. The phenomenon of alien plant invasions may therefore be an emerging problem in arid parts of South Africa. Established invasive plants have spread more quickly in arid than in wetter areas of South Africa (Wilson et al. 2007). Certainly, plant invasions have been much less studied in arid parts of South Africa, and information on the processes of invasion in such areas is urgently needed (Richardson & van Wilgen 2004).
The low density of trees, and therefore subcanopy microsites, in semi-arid savannas results in strongly clumped dispersal of seed by birds (Dean et al. 1999). This clumping increases the probability that even small individuals of alien plant species present at low density could be detected.
We therefore undertook this study of the early stages of a plant invasion to test the following hypotheses:
1invasion by alien fleshy-fruited trees and shrubs into semi-arid savanna is aided by the presence of host trees;
2propagule pressure determines invasion probability; and
3alien host plants are better facilitators than native host species of invasion by alien fleshy-fruited species, i.e. the conditions exist for invasional meltdown.
Materials and methods
The study was conducted in and around the Diskobolos Military Base (Aleksanders Fontein), 2 km south of Kimberley in South Africa's Northern Cape province (28°47′ S, 24°47′ E, 1198 m a.s.l., Fig. 2). The climate is semi-arid and the mean annual rainfall is 414 mm, most of it from thunder storms in late summer. Mean daily temperatures range from 33 °C to 18 °C in January, and from 19 °C to 3 °C in July (South African Weather Bureau 2003). The military base was established in 1967, and many trees of several species, all now well established, were planted along the streets to provide shade and decoration at this time. The military and associated personnel also maintain gardens in which many more plants are cultivated, most of them alien or extra-limital species (i.e. species indigenous to South Africa but outside their natural distribution ranges).
The savanna vegetation surrounding the military base is classified as Kimberley Thorn Bushveld (Low & Rebelo 1996). The dominant native tree species is Acacia tortilis (Fabaceae: Mimosoideae). Associated with this are native bird-dispersed plants including Rhus lancea (Anacardiaceae) that have potential to overtop A. tortilis, and various subcanopy shrubs, lianas and herbs. Tree cover varies between 5% and 15% at the hectare scale (based on estimates made from aerial photographs). The planting of alien trees in adjacent areas (e.g. at the nearby Kimberley Airport) is minimal, thus making the military base and surrounding savanna ideal for a study of plant invasion from a single source. There are also three long-established invasive plant species absent from the base but already present at the landscape scale in the surrounding vegetation. These are pod-bearing Prosopis sp. (Fabaceae: Mimosoideae; mesquite) (hybrids of P. glandulosa and several other alien Prosopis taxa), and the fleshy-fruited species Opuntia ficus-indica (Cactaceae) and Atriplex semibaccata (Chenopodiaceae).
Inside the military base, all adult fleshy-fruited plants of species not occurring naturally in the Kimberley area, and that would represent potential propagule sources for invasion in the surrounding savanna, were identified and mapped. Outside the base, 181 sites, each comprising a canopy tree and an adjacent plot in herbaceous vegetation, were surveyed. The information recorded included (i) tree species, (ii) height (estimated to the nearest metre), (iii) minimum and maximum canopy radius, (iv) geographical coordinates, (v) soil type (classified as sand, loam or calcrete (also known as caliche)), and (vi) presence or absence of any obvious anthropogenic or natural disturbance, such as mining, construction, paths, or burrows of porcupines (Hystrix africaeaustralis) and aardvarks (Orycteropus afer). The fleshy-fruited plant species (all life-forms) occurring under each tree were recorded in seven height classes (0–0.05 m, 0.05–0.1 m, 0.1–0.2 m, 0.2–0.40 m, 0.40 m–canopy, in canopy, emergent). The open plot (5 × 5 m) was placed 20 m from the canopy of each sampled tree in the direction least affected by neighbouring trees, and was surveyed in the same way for fleshy-fruited plant species. The spatial positioning of the surveyed plants is given in Fig. 2. All surveys were undertaken during April 2006, following autumn rain, when the probability of finding young seedlings was greatest.
We considered three main questions related to the processes and hypotheses outlined. First, to what extent is the distribution of seedlings of a species defined by where adult plants are (processes 2 and 5 on Fig. 1; hypothesis 1)? Secondly, what were the main determinants of seedling species richness and abundance, in particular how did the relative proportion of aliens to natives vary with distance from the source of most of the aliens (the military base) (processes 2 and 5 on Fig. 1; hypothesis 2)? Thirdly, does the identity of the host tree affect the assemblage composition of the fleshy-fruited plants species (process 3a, 3b and 4c on Fig. 1; hypothesis 3)?
To examine the effect of distance on the probability of any particular species dispersing to a site, we modelled species abundance under a host tree as explained by the presence of disturbance, the log of the Euclidean distance to the closest mature plant, and the log of canopy extent using a generalized linear model with negative binomial errors. As a measure of canopy extent we used the product of the canopy radius and canopy height as this was the best explanatory variable for patterns of species richness (see results), and other measures of canopy extent, e.g. canopy volume, gave qualitatively similar results. The minimum adequate model was arrived at by step-wise deletion from the maximal model. As the species abundance data were often highly skewed (a few sites having many seedlings), generalized linear models with negative binomial errors were used (glm.nb in Venables & Ripley 2002). Subcanopy individuals that were the same species as the host tree were excluded, as were species adapted for dispersal by wind. Therefore the analyses only explore individuals that were dispersed by birds or other animals. For native trees, the distance to the centre of the military base was measured to test whether proximity affected their frequency.
Patterns in species richness were investigated by fitting a generalized linear model with Poisson errors. The minimal adequate model was arrived at by step-wise deletion from the maximal model (with a cut-off of P(×) = 0.05). The percentage deviance explained by each significant factor was determined by the change in explained deviance when the factor was dropped from the minimum adequate model (Lobo et al. 2002). To explore the spatial structure of the variation, a minimum adequate trend surface model was produced (Legendre & Legendre 1998). All significant environmental and spatial terms were then added in a full model, and the spatially structured components of the environmental variables were assessed by comparing the percentage deviance explained by the factor in a model without the trend surface terms with the deviance explained by the factor in a spatially implicit trend surface model. In most cases the residuals were not over-dispersed, but, as a check of the robustness of the results, the models were re-run using a quasi-Poisson error structure.
The process was repeated for alien species richness and native species richness separately. To test for differences between germination and establishment, we re-ran all analyses excluding records of plants < 0.1 m tall, which we assumed to represent seedlings. Finally, to explicitly test for differences between the relative species richness of natives and aliens found under the hosts, we used generalized linear models with binomial errors.
To ascertain whether fleshy-fruited plant assemblages were different when they occurred beneath native (i.e. A. tortilis) or alien (i.e. Prosopis) host trees, we constructed similarity matrices of log abundance data using a Bray-Curtis similarity measure, and then used an analysis of similarities (anosim) to detect significant differences. These comparisons were carried out on data sets from which seedlings had been excluded (i.e. subcanopy plants < 0.1 m in height), as we were interested in the final species composition beneath trees and not species that germinate but do not establish. Those trees for which removal of seedlings left no subcanopy fleshy-fruited plants could not be analysed and were excluded. We also analysed similarities in subcanopy alien species composition, using the same methods as above, to ascertain whether alien species assemblages were significantly different as a function of native and alien host tree, canopy extent and distance from base. These analyses were done using PRIMER vs. 5 (Clarke & Gorley 2001, 2006).
importance of host trees as establishment sites
The presence of host trees appeared to be essential for invasion by alien fleshy-fruited trees and shrubs (hypothesis 1). While 19 native and 12 alien fleshy-fruited species were found growing beneath the tree canopies, 11 native (but no alien) fleshy-fruited species were recorded in the plots outside the tree canopies (Tables 1 and 2; the ratio of alien to native species was higher under the canopy than in the open, Fisher's test P = 0.018; Thesium hystrix and one Asparagus species were found in the outside plots but not under the canopy, so the total recorded species richness was 33). Of the plots outside the canopy, 62 contained no fleshy-fruited species, 97 had one species, 21 had two species and one had three species; of those species found outside the canopy only Lycium cinereum, Lycium hirsutum and Pollichia campestris were present in more than three plots. In contrast, under the canopy half the species were present under more than 10 host trees, and median species richness per host tree was five (range 0–14). There was no significant correlation between species richness under tree canopies and in the adjacent grassy plots (Spearman's rank correlation, P = 0.20).
Table 1. Alien or extra-limital fleshy-fruited species found in the study area. Species are ranked according to a qualitative assessment of invasion probability. Species that have mature individuals in the field and are frequent in natural vegetation are most likely to become invasive. Those species not found outside the base were ranked according to decreasing propagule pressure (more individuals of a species should provide a stronger test of its invasion potential). The distance relationship is a test of whether distance from the nearest mature plant (or the centre of the base if no mature plants were recorded) significantly affects the abundance of the species at a site (i.e. whether there is evidence of dispersal limitation). The combined list of P-values for native (Table 2) and aliens (this table) was corrected for multiple comparisons using the false detection rate method (P-adjust function in R). Columns marked NA denote where tests could not be performed (e.g. the species was not found outside the base, or only one or two individuals were recorded)
There were positive relationships between site occupancy and seedling abundance for both small (< 0.1 m in height) and larger individuals (Fig. 3a,b, P < 0.01). While these relationships did not differ significantly between natives and aliens, the most abundant native fleshy-fruited tree species (Rhus lancea) dominated assemblages of seedlings < 0.1 m in height, and consequently had a large influence on the analysis (Cook's distance > 1, Fig. 3a). If small Rhus lancea seedlings were excluded from the analysis, site occupancy, relative to seedling abundance, was significantly lower for alien than native species, indicating a higher degree of clumping in deposition of seed of aliens, potentially linked to the propagules arising primarily in the military base.
propagule pressure and invasion probability
The survey inside the military base revealed that 39 fleshy-fruited plant species (573 individuals) that do not occur naturally in the Kimberley area were in cultivation, most as ornamentals and shade trees (e.g. Melia azedarach, Schinus molle, Rhus pendulina) or for fruit (e.g. Punica granatum, Prunus persica, Prunus armeniaca). Only nine of these species were recorded in the surrounding savanna. Of these, two (Olea europaea ssp. africana and Rhus pendulina) are native to South Africa but not to the local area. The number of seedlings outside the base was strongly correlated with the number of mature plants found inside the base (Fig. 4, P < 0.01), and no species that had fewer than five individuals inside the base was found outside it. Seed size of alien species explained neither distance dispersed nor number of individuals found outside the base. These observations are in agreement with hypothesis 2, namely that propagule pressure determines invasion probability.
We also used size-class histograms (see Fig. S1 in Supplementary Material for examples) to assess the probability of invasion for the alien species encountered (Table 1). Size-class distributions differed considerably among species. Whereas all the native species were represented as mature individuals, two of the alien tree species were present only as seedlings or saplings.
When investigating how far the alien species had dispersed from the military base, we included disturbance and canopy extent as covariates because trees sampled further away tended to have larger canopies (P < 0.05) and more of the sites closer to the base were disturbed (P < 0.001). There was, however, no significant correlation between distance to the military base and host plant identity (P = 0.35) or soil type (P = 0.51 for sand, P = 0.55 for calcrete), so these factors were not included here. Three types of distribution patterns for fleshy-fruited species were identified (Fig. S2, Table 1): density higher closer to the military base or adult plants; density higher further from the military base; and density independent of distance from the base (most native species and some alien species). The alien species Olea europaea ssp. africana, Rhus pendulina and Punica granatum were strongly restricted to sites closer to mature conspecifics. Schinus molle and Melia azedarach seedlings also tended to occur close to adult trees (Table 1), but several mature individuals occurred outside the military base and so it would be problematic to infer the species was only dispersing to areas close to the base. Of the natives, only an Asparagus species and Ehretia rigida were affected by proximity to the base, and, in these cases, they were more abundant further from the base (Table 2). There were no significant trends for the other species.
host effect (i.e. native or alien) on species richness patterns
The host species in the study represented similar geometric shapes in terms of canopy height and radius (F10,175 = 0.975, P = 0.47). Similarly, the relationship between height and radius was not significantly different between the Prosopis and Acacia tortilis trees sampled (F2,155 = 3.01, P = 0.052). Therefore, any differences in host suitability should be ascribed to properties of the host (e.g. differences in canopy structures or shade) and not larger areas per se.
Subcanopy fleshy-fruited species richness declined with distance from the military base for Prosopis sp. but was independent of distance for Acacia tortilis (Fig. 5, Table 3a). The parameter estimates are such that across all observed distances from the base the subcanopy species richness of Prosopis is predicted to be less than that for Acacia tortilis. Total subcanopy fleshy-fruited species richness under Prosopis sp. and Acacia tortilis increased with canopy extent (Fig. 6), and was lower in more disturbed sites (Table 3a).
Table 3. (a) Minimal adequate model to explain total species richness under the canopy (log-link function). The estimates and standard errors shown are for each factor level, the P-value and percentage deviance are for the factor as a whole. Total percentage of deviance explained can sum to greater than the explained percentage deviance because of colinearity between the factors (e.g. trees further away from the military base tended to be slightly smaller and so in the absence of distance from the base, the parameter for canopy extent in the model will explain a portion of the variance attributable to distance). Results were qualitatively similar if all host plants were included in the analyses. Distance from the base and the volume of the subcanopy were both log-transformed. Null deviance (for model without spatial component): 255.30 on 149 degrees of freedom. Residual deviance: 149.46 on 144 degrees of freedom. (b) Minimal adequate model to explain native species richness under the canopy (log-link function). Null deviance (for model without spatial component): 163.25 on 149 degrees of freedom. Residual deviance: 107.84 on 146 degrees of freedom. (c) Minimal adequate model to explain alien species richness under the canopy (log-link function). Null deviance (for model without spatial component): 271.48 on 149 degrees of freedom. Residual deviance: 172.21 on 143 degrees of freedom. (d) Minimal adequate model to explain the relative proportion of native to alien species richness under the canopy (logit-link function). Null deviance (for model without spatial component): 189.26 on 153 degrees of freedom. Residual deviance: 145.87 on 149 degrees of freedom
Model without spatial component
Trend surface model
Base model is host =Acacia tortilis, no disturbance
Overall these four factors (distance, canopy extent, host identity and disturbance) explained 41.6% of the model deviance, and the data were well described by a Poisson distribution (deviance per degree of freedom of 1.04). Canopy extent explained more variation than canopy radius (total model deviance explained of 37.8%), canopy height (39.0%), or canopy volume (40.5%), and if the various measures of the canopy dimensions were included in the maximal model only canopy extent remained in the minimum adequate model. When a trend surface model was considered, disturbance was no longer a significant factor, but the percentage variance explained by the other factors did not change much. Soil type (loam, sand and loam, or calcrete and loam) did not have any significant effects.
When only natives were considered, the only significant effects were host, canopy extent and disturbance (Table 3b). Moreover, disturbance was not significant in the trend surface model, so the native species richness is in fact mostly described here by the size and species identity of the host tree. This qualitative result was the same if all host trees were included.
Species richness of aliens was explained by canopy extent, soil type and an interaction between host and distance from the military base (Table 3c). There were no significant differences in the proportions of species that were alien under Prosopis sp. and Acacia tortilis (P (×) = 0.077), and similarly there was no effect of host plant if all hosts were considered (P (×) = 0.174). This is contrary to the prediction that non-native host plants would facilitate invasions (hypothesis 3). However, the proportion of aliens to natives in the subcanopy was greater closer to the base (Fig. 5; providing support for the role of propagule pressure in the invasion process (hypothesis 2)), greater on sandy sites and lower on calcrete sites, and also increased with canopy extent (Table 3d). Much of the explanatory power of these factors was spatially auto-correlated, with only canopy extent significantly affecting the proportion of aliens to natives in the trend surface model.
The same minimum adequate models and qualitative results were obtained for all three species richness measures and for the proportion of native species to alien species when only subcanopy fleshy-fruited individuals greater than 0.1 m in height were considered.
A total of 158 trees were used in the comparison of subcanopy assemblages beneath A. tortilis and Prosopis, after seedlings were removed and trees with no other fleshy-fruited subcanopy plants excluded. Subcanopy assemblages were significantly different beneath A. tortilis and Prosopis canopies (Global statistic R = 0.143, P < 0.01).
An analysis of the species contributing most to the observed differences revealed that the differences were primarily due to changes in relative abundance of native fleshy-fruited species (Table 4). Most species were more common under A. tortilis, although Rhus lancea was more common beneath Prosopis, and Lycium cinereum had comparable abundances beneath the two. Relative abundance of subcanopy fruited species differed between the two species. Beneath A. tortilis (in order of declining abundance), P. campestris, Diospyros lycioides, R. lancea, Ziziphus mucronata and L. cinereum were the most common species. Beneath Prosopis the order was R. lancea, P. campestris, D. lycioides and L. cinereum.
Table 4. Subcanopy species contributions to differences in assemblages beneath A. tortilis and Prosopis host trees. Numbers in bold indicate higher abundance compared between the two host-tree species
Average abundance beneath A. tortilis
Average abundance beneath Prosopis
Dissimilarity/ standard deviation
Comparison of only the alien component of fleshy-fruited species found no significant differences between host tree species or distance from military base, although their assemblage composition differed with canopy extent (P < 0.01). Here, Celtis sp. and Schinus molle were far more common beneath large trees, and Echinopsis spachiana was more common beneath small trees. Again, this provided no evidence in support of alien host plants facilitating invasions more than native host plants (hypothesis 3).
Birds facilitate the movement of alien plant species from gardens into a variety of natural and anthropogenic ecosystems (Glyphis et al. 1981; Ferguson & Drake 1999; Dean & Milton 2000; Gimeno & Vila 2002; Gosper et al. 2005; Buckley et al. 2006). In so doing they assist plants to overcome an important barrier to invasion, namely seed dispersal (Richardson et al. 2000b). By dispersing plants in a directed way to potentially suitable microsites, birds may also help invaders over various additional obstacles to invasion. Thereafter, the components of invasion risk (probability and impact) are more difficult to predict because they vary with species, ecosystems, interactions and feedbacks.
Our study tested three hypotheses, namely: that invasion is aided by host trees (for which we found strong support); that propagule pressure determines invasion probability (again strongly supported); and that alien host plants are better facilitators of alien fleshy-fruited species invasions than are native hosts (not supported).
importance of host trees as establishment sites
Although the evolution of fruit for directed dispersal may only be a secondary adaptation (Mack 2000), directed dispersal to suitable sites can reduce seedling mortality and consequently raise fitness (Howe & Smallwood 1982; Wenny 2001), suggesting that evolution of fruit confers the advantage of arrival in suitable establishment sites (Howe & Smallwood 1982). Dispersal to sites in the savanna was mainly by birds or bats, and most fleshy-fruited plant species established only in microsites beneath tree canopies. Performance of fleshy-fruited seedlings may be better in subcanopy sites than in the open (Milton et al. 1997), although this does not apply to all fleshy-fruited plants when they reach maturity (Seymour 2006), nor are shaded sites a prerequisite for successful establishment of all fleshy-fruited plants (Dean & Milton 2000). Nucleation then arises because seeds of fleshy-fruited alien plants are directed to subcanopy sites by frugivores such as birds or bats. Dispersal of seeds by frugivorous birds is generally clumped beneath perch sites (Snow 1976; Milton et al. 1997; Holl 1998). We assume that the observed association of alien fleshy fruited plant species with tree canopies was primarily the result of dispersal by birds, because vervet monkeys (Cercopithecus aethiops), while occurring sporadically in this area, are not sufficiently common to have a significant impact on seed rain to canopy sites (W. R. J. Dean, personal observations).
Seeds of alien plants dispersed to the savanna ranged in diameter from 1 mm (Ficus carica) to 8 mm (Melia azedarach), within the range taken by the most common frugivorous birds in this system, namely Red-faced Mousebirds (Urocolius indicus), Acacia Pied Barbets (Tricholaema leucomelas), African Red-eyed Bulbul (Pycnonotus nigricans) and Cape Glossy Starlings (Lamprotornis nitens) (Seymour 2006). Some alien fruits also resembled those of native plants in size, colour and texture. For example, the fruits of Celtis sp. (yellow and leathery) resemble those of the native species Ziziphus mucronata and Grewia flava, while those of Schinus molle and Rhus pendulina are very similar to those of local Rhus species. As monkeys are uncommon, and as there are no indigenous plant species with large dispersal units resembling those of Citrus sp. or Maclura pomifera, it is unlikely that indigenous birds would be able to open up these fruits and disperse the seeds.
We have shown that, in the absence of host trees which attract birds and provide subcanopy sites where conditions are favourable for seedling establishment, alien fleshy-fruited plant species would either fail to be dispersed from gardens into savanna, or would fail to establish there. Host trees are therefore essential for invasion of savanna by fleshy-fruited plants.
propagule pressure and invasion probability
Propagule pressure is a crucial mediator in the interplay of the numerous factors that determine the success of an invasive species (review in Richardson & Pyšek 2006).
For species that potentially had a high propagule pressure but are not invading, there were easily identifiable barriers to dispersal and establishment, namely: low or non-viable seed production; absence of suitable dispersers, in particular for Maclura pomifera, Poncrius sp. and Citrus sp.; absence of suitable subcanopy establishment sites; or a lack of conditions required to break dormancy, e.g. Rosa damascena and Pyracantha sp. may have failed to invade because conditions were never cold and moist enough to promote germination (Anonymous 2005). However, the reasons for some of the other species, e.g. Ligustrum lucidum, not establishing outside the military base are unclear.
The presence and relative abundance of the various alien plant species cultivated in the military base probably reflects their net performance over the past 40 years. Although most of the gardens were irrigated, species suited to the ambient environmental conditions were probably planted more frequently or survived better. Consequently, the most common species on the base may be those best adapted to conditions in the surrounding vegetation. Smaller-seeded plants are also expected to be more efficiently dispersed (Levey 1987), but this effect was not seen here, perhaps because a limited suite of species can establish in this semi-arid environment.
Despite these exceptions, our study provides clear evidence that an abundant seed source (in the presence of suitable seed dispersers and host trees) will increase the probability of invasion of a natural ecosystem by alien species.
Our results show the net performance of the various alien species over several decades of fluctuating weather conditions. Size-class distribution for alien trees indicated that whereas some (Celtis sp., Rhus pendulina, Schinus molle) could already be objectively classified as invasive (Richardson et al. 2000b) on the basis of widespread distribution of seedlings and ability to establish and reproduce in undisturbed savanna, others (Brachychiton populneus, Ficus carica) were well dispersed but have yet to reach reproductive maturity. If the fire regime at the site was to change or there were several wet years, these species may recruit, but at present they have a low probability of invading this habitat type.
The relationship between seedling abundance and site occupancy for small seedlings of alien plants indicated that invasion of the savanna site sampled was still in a relatively early stage (Fig. 3). Large numbers of small seedlings occurred at a few sites but alien seedlings were not as well dispersed as those of natives. This was probably because they originated from the military base alone, or a few point sources for those species where individuals had reached reproductive maturity in the savanna.
Lawes et al. (2006) found that alien plants that were abundant locally tended to become abundant at landscape level and could be considered high-risk invaders. On the basis of this framework, Celtis sp., Rhus pendulina and Schinus molle pose the biggest invasion risks in the area and all three species require coordinated management action.
extra-limital indigenous invaders
Two of the top five savanna invaders identified in this study, R. pendulina and O. europaea ssp. africana, are indigenous to South Africa and naturally occur within 200 km of the study site, R. pendulina in riparian woodland along the Orange River, and O. europaea ssp. africana on dolomitic soils on the Ghaap Plateau (Low & Rebelo 1996). As hypothesized, it would appear that the propagule pressure from plantings in the military base also enabled these species to invade Acacia savanna despite the habitat differences. Moreover, the most invasive species, Celtis sp., is probably a hybrid between the indigenous Celtis africana that occurs in forest pockets in higher altitude grassland to the east of Kimberley, and C. australis or C. sinensis, both of which are used as garden subjects. Hybrid seedlings have previously been reported from the suburbs around Pretoria in the Gauteng province of South Africa (Coates Palgrave 2002).
native vs. alien hosts and the potential for ecosystem transformation
Our hypothesis that alien host plants are better facilitators of alien fleshy-fruited species invasions than are native hosts was rejected. Both abundance and species richness of fleshy-fruited species were lower beneath alien Prosopis spp. than beneath indigenous Acacia tortilis, suggesting that, at this site, ‘invasional meltdown’ (sensu Simberloff & Von Holle 1999) is not implicated and that alien fleshy-fruited plants can infiltrate natural savanna woodland more easily than woodlands invaded by mesquite. In addition, as both native and alien fleshy-fruited species were more common beneath native host trees, replacement of native by alien hosts would see a decline in over- all species richness. Dean et al. (2002) also found that fleshy-fruited species occurred at lower densities beneath Prosopis than Acacia species. This cannot be explained by different longevities of Prosopis sp. and Acacia tortilis because they have similar potential life spans of 100–200 years (Prince & van der Jeugd 1993; Flinn et al. 1994). Lower species richness below Prosopis might be because Acacia species offer better perches for frugivorous birds (Dean et al. 2002), or because Acacia species provide subcanopy environments more suitable for seedling germination and survival.
Although alien host trees do not promote invasion of alien fleshy-fruited plants more successfully than do native hosts, subcanopy plant assemblages differed significantly among host species. Alien hosts and subcanopy invaders may also respond differently to disturbances or alter a variety of species interactions, leading to unpredictable changes in vegetation dynamics. For example, alien plants that can invade relatively undisturbed ecosystems have the potential to transform them if they can out-compete indigenous plants for essential resources (Flores-Flores & Yeaton 2000), change successional dynamics of local savannas (Archer et al. 1988; Dean et al. 1999), or disrupt disperser or pollination mutualisms (Traveset & Richardson 2006). Whereas in natural savanna only Rhus lancea and occasionally Ziziphus mucronata are able to overtop their Acacia hosts, it is likely that the potentially taller alien species Schinus molle and Celtis sp. would do so more frequently, increasing mortality rates of native trees.
Phenologies of native fleshy fruiting species in the Kalahari are timed to minimize competition for dispersers and fruit. Fruit of Rhus spp. ripens in spring (September) followed by D. lycioides and E. rigida in summer (November–January), and later in autumn and winter (March–August) by G. flava and Z. mucronata (Coates Palgrave 2002), whereas Asparagus, Lycium, Pollichia, Solanum and Thesium spp. can fruit more than once annually following opportunistic flowering responses to large rainfall events (Milton et al. 2004). Alien species might out-compete native species for dispersers or change dispersal distances if they supply more fruit and overlap in fruiting phenology (Traveset & Richardson 2006). In addition, differences in subcanopy assemblage composition are likely to have implications for pollinators, invertebrate herbivores and dispersers.
Alien species also have the potential to change the intensity of fires beneath tree canopies. Fire often kills large trees, and changes in understorey species composition and leaf biomass are likely to change the way the fire burns (e.g. temperature, position and duration). Seymour (2006) found that fire decreased cover of fleshy-fruited plants in Kimberley savanna, but that some species could resprout. Differences between alien and native trees in longevity, flammability and resprouting ability may affect the potential of alien invaders to transform Kalahari savanna.
It is clear that the system will be changed by the infiltration of alien (including indigenous extra-limital) species. Invasive species can potentially change dispersal dynamics. Indigenous fleshy-fruited species in this area compete for dispersers, as there is almost always a surfeit of fruit on subcanopy plants, so alien species are likely to increase that competition. Of the native species recorded, only R. lancea overtops the shade-intolerant A. tortilis. However, at least two of the alien species, S. molle and Celtis sp., also have this potential, and so the invasion may increase the likelihood that a host tree is over-topped, shaded and killed. On a positive note, the management of invasive species can concentrate on the subcanopy environments, as these are the only sites where the fleshy-fruited alien species can establish.
This study was funded by the DST-NRF Centre of Excellence for Invasion Biology. S.P. acknowledges postdoctoral funding from Stellenbosch University. C.L.S. was supported by the German Ministerium für Bildung und Wissenschaft (BMBF) under project number 01C0024 (BIOTA – southern Africa) and the Gordon Sprigg Scholarship. We thank the DST-NRF Centre of Excellence at the Percy FitzPatrick Institute of African Ornithology for use of the Benfontein Research Centre as our base, and Eric Herrmann for general assistance. The Commanding Officer Maj. J.A. Myburgh of the Kimberley Military Base is thanked for allowing access to the base and surrounds. Ruan Veldtman and John Field provided statistical advice.