Invasive mutualisms and the structure of plant–pollinator interactions in the temperate forests of north-west Patagonia, Argentina

Authors


C.L. Morales (fax +54 2944 422111; e-mail cmorales@crub.uncoma.edu.ar).

Summary

  • 1Alien species may form plant–animal mutualistic complexes that contribute to their invasive potential. Using multivariate techniques, we examined the structure of a plant–pollinator web comprising both alien and native plants and flower visitors in the temperate forests of north-west Patagonia, Argentina. Our main objective was to assess whether plant species origin (alien or native) influences the composition of flower visitor assemblages. We also examined the influence of other potential confounding intrinsic factors such as flower symmetry and colour, and extrinsic factors such as flowering time, site and habitat disturbance.
  • 2Flowers of alien and native plant species were visited by a similar number of species and proportion of insects from different orders, but the composition of the assemblages of flower-visiting species differed between alien and native plants.
  • 3The influence of plant species origin on the composition of flower visitor assemblages persisted after accounting for other significant factors such as flowering time, bearing red corollas, and habitat disturbance. This influence was at least in part determined by the fact that alien flower visitors were more closely associated with alien plants than with native plants. The main native flower visitors were, on average, equally associated with native and alien plant species.
  • 4In spite of representing a minor fraction of total species richness (3.6% of all species), alien flower visitors accounted for > 20% of all individuals recorded on flowers. Thus, their high abundance could have a significant impact in terms of pollination.
  • 5The mutualistic web of alien plants and flower-visiting insects is well integrated into the overall community-wide pollination web. However, in addition to their use of the native biota, invasive plants and flower visitors may benefit from differential interactions with their alien partners. The existence of these invader complexes could contribute to the spread of aliens into novel environments.

Introduction

Invasive species affect many ecosystem services, including pollination of plants by animals (Chittka & Schurkens 2001; Memmott & Waser 2002). Invasive alien plants establish direct interactions with local pollinators, either native or alien to the sites they invade (Morales & Aizen 2002; Olesen et al. 2002), which guarantee the maintenance and eventual spread of populations via seed (Parker 1997). Alien plants can also establish indirect interactions with native plants through pollination, which may set the stage for pollinator-mediated competition (Chittka & Schurkens 2001; Brown et al. 2002). The importance of these indirect interactions depends, to a large extent, on the similarity of the flower visitor assemblages of alien and native species.

The identification of key factors influencing the structure of plant–pollinator interactions in communities where aliens have established represents a first step in evaluating the impact of invasions on plant communities. Most studies assessing the impact of alien plants and pollinators have focused on the relationship either between a single alien flower visitor species and its associated plants (Dafni & Shmida 1996) or between a single alien plant species and its associated flower visitors (Parker 1997; Barthell et al. 2001). However, recent research has scaled up to include all the interactions occurring in substantial parts of plant–pollinator communities. This ‘pollination web’ approach has provided new and more general insights into how alien plants and flower visitors integrate and how they spread into native communities (Memmott & Waser 2002; Morales & Aizen 2002; Olesen et al. 2002).

A logical disadvantage of focusing at the community level is the loss of precision regarding the actual contribution of every single potential pollinator species to the pollination of each plant species. Most studies on plant–animal interactions at this level do not distinguish between flower visitors and pollinators and therefore ignore, beyond measures of visit frequency, most biological details of the interaction (Hingston & McQuillan 2000; Olesen et al. 2002; Potts et al. 2003; but see Memmott & Waser 2002; Vázquez & Simberloff 2003). Here we define flower visitors as those animals, mainly insects, that visit flowers seeking nectar or pollen. Most of them make contact with sexual parts, although efficient pollen transfer may vary greatly among species. However, recent advances showed that visit frequency, one of the two important components of pollination efficiency, might closely predict the pollination importance of a given flower-visiting species (Vázquez & Aizen, in press; D.P. Vázquez, unpublished results).

One basic question that needs to be addressed is whether plant species of different origin (alien or native) differ in their associated flower visitor fauna. A more in-depth analysis should tease apart the factors structuring plant–pollinator communities. After all, alien and native species may differ in many aspects beyond their geographical origin that could account for differences in flower visitor assemblages. Flower visitor assemblages could be influenced by intrinsic factors, such as different floral traits that will determine which subset of all the available flower visitors will visit and pollinate flowers (Dicks et al. 2002). They can also be influenced by extrinsic factors, such as the characteristics of the site where a plant species occurs, and the time when it blooms. These may relate directly to pollinator abundance and determine the high temporal and spatial variation of most pollinator assemblages (Herrera 1988, 1995; Hingston & McQuillan 1999, 2000; Aizen 2001; Hambäck 2001; Potts et al. 2003).

Among intrinsic factors, some floral features, such as corolla colour or flower symmetry, have traditionally been associated with certain flower visitor types (Faegri & Van der Pijl 1979). For instance, bees visit a wide range of flowers, but in general, they favour yellow or purple flowers. Bilateral symmetry (i.e. zygomorphism) is also a common feature of bee flowers, particularly associated with pollination by bumblebees and other large bees (Proctor et al. 1996). Among extrinsic factors, in highly seasonal climates, although a few flower visitors are active year round (e.g. Aizen 2003), most flower visitors forage on flowers for just a few weeks each year (Bronstein 1995). Thus, the temporal overlap between flowering phenology and the period of pollinator activity may partly shape flower visitor assemblages. Regarding spatial variability, individuals of the same species growing in different habitat types or under different microclimatic conditions could differ greatly in the composition of their flower-visiting fauna (e.g. Herrera 1988, 1995), whereas different co-occurring species may share a similar array of flower visitors in spite of differences in their morphology, colour or rewards (Feinsinger 1983).

In particular, changes caused by anthropogenic habitat disturbance could affect the number, identity and behaviour of flower-visiting animals (Aizen & Feinsinger 1994; Klein et al. 2002; Morales & Aizen 2002; Vázquez & Simberloff 2003). Disturbances like fire, overgrazing and forest clearing may affect flower visitors in different ways: (i) they could reduce both sites for nesting and the availability of flowering plants (Aizen & Feinsinger 2003; Potts et al. 2003); (ii) they may favour the flowering of understorey species (Ghazoul 2002) and an increase in flower visitor abundance; and (iii) they can mediate an association between alien plants and animals by fostering invasion of both mutualistic partners (Barthell et al. 2001; Morales & Aizen 2002).

Invasive plants are one of the most important threats to native biota in the temperate forests of NW Patagonia, Argentina (APN 1986, 1992; Aizen et al. 2002). These forests are characterized by a high incidence of plant–pollinator mutualisms (Smith-Ramirez & Armesto 1994; Aizen & Ezcurra 1998). Thus, a detailed characterization of which flower-visiting species interact with each plant species, and how frequently, could help us not only to describe the structure of this mutualistic web but also to elucidate the potential effects of alien plants and insects on pollination interactions. In particular, we were interested in assessing the impact of plant species origin on the structure of this pollination web, and which intrinsic and extrinsic factors could account for a potential association between alien plants and flower-visiting insects. To our knowledge, only two studies have analysed the integration of alien plants and flower visitors in pollination webs: Memmott & Waser (2002) analysed a data base of 1429 animal species visiting flowers of 456 plant species in small farms, orchards and hedgerows in central USA, and Olesen et al. (2002) compared pollinator networks from two widely separated oceanic islands belonging to the Azores and Mauritius archipelagos, respectively. The novel contribution of our study is the formal incorporation of factors that potentially covary with species origin in order to explain the patterns of plant–pollinator interactions in a community composed of alien and native mutualists.

We investigated whether the origin of the plant species (alien vs. native) influences the diversity and composition of their flower visitor assemblages. In particular, we asked whether alien plants are preferentially associated with alien flower visitors. We then tested whether the influence of species origin persists after accounting for differences in corolla colour, flower symmetry, flowering time, site and disturbance. In particular, we asked whether plant species origin per se affects the composition of the flower visitor assemblage, beyond the well-known association between alien species and disturbed habitats.

Materials and methods

study area and study sites

We carried out fieldwork in Nahuel Huapi National Park and surrounding areas in NW Patagonia, Argentina (c. 41°S) during the 2000–01 austral flowering season. We chose four sites along a 50-km transect along a west–east gradient of decreasing precipitation (Puerto Blest, Llao Llao, Cerro Otto and Challhuaco) representing, respectively, Valdivian temperate rain forest (780 m a.s.l.), mixed forest of Nothofagus dombeyi (Mirb.) Blume and A. chilensis (800 m a.s.l.), Austrocedrus chilensis (D.Don) Flor et Boutleje forest (870 m a.s.l.), and high altitude forest of Nothofagus pumilio (Poepp. Et Endl.) Krasser (990 m a.s.l.). At each site, we set up two contiguous sampling plots (c. 2 ha each) characterized by contrasting intensities of habitat disturbance (hereafter disturbed vs. undisturbed habitat). Disturbance involved an opening of the forest canopy due to either clear cutting or fire. A more detailed description of the sites and types of human disturbances occurring at each site can be found in Morales & Aizen (2002).

study species

We studied the flower visitor assemblages of a subset of the plant community comprising the 13 alien and 15 native animal-pollinated species that were most abundant in terms of number of individuals and flower abundance, and probably the most relevant in determining the structure of this plant–pollinator web (Vázquez & Aizen in press). Many species were present in both disturbed and undisturbed habitats within a site, and some of them occurred in more than one site. Sites, and the two habitat units within each site, varied in the diversity of alien and native plant species and therefore in the number of species sampled. For example, in Puerto Blest, where a published flora is available (Brion et al. 1988), we sampled 8 alien and 10 native species, which represented 32% and 11% of the alien and native animal-pollinated flora, respectively. In the four sites, richness and abundance of alien plant species in the disturbed habitat were higher than in the undisturbed habitat, particularly because many alien species did not occur in the adjacent undisturbed habitat (C. Morales, unpublished data).

field methods

Throughout the blooming period of each species, we recorded the identity and number of visits made by diurnal flower visitors to flowering branches (shrubs and treelets) or flower patches (herbs) during 15-minute censuses and used these to compute the per-flower visit frequency per census (number of visits flower−1 15 minutes−1). We considered the composite flowering head of species in the Asteraceae (see Appendix S1 in Supplementary Material) as well as the inflorescence of Trifolium repens L. as the functional unit of visitation (e.g. Aizen & Feinsinger 1994; Barthell et al. 2001). We made the flower visitor censuses between 09.00 and 18.00, alternating observations between the disturbed and undisturbed habitat within each site. Different sites were sampled on different days and each site was sampled at least once a week. We conducted a total of 1639 censuses on 73 days.

With the exception of those species that were easily recognizable in situ, insects were captured and determined to the lowest taxonomic category possible, using the keys of Edwards (1930) for Nemestrinidae (Diptera), Shannon & Aubertin (1933) for Syrphidae (Diptera), and Moure (1964) and Goulet & Huber (1993) for Hymenoptera. Any insect that could not be identified to species, genus or even to family level was morphotyped. This is still useful for the purpose of characterizing and quantifying flower visitor communities (see also Memmott & Godfray 1993; Oliver et al. 2000). Voucher specimens have been kept at Laboratorio Ecotono, Universidad Nacional del Comahue and at the Museo de Ciencias Naturales Bernardino Rivadavia. As only four out of 110 identified insect species were alien, we are confident in the precision of our classification in relation to the origin of flower visitors (alien or native) in the field.

data analysis

We compared the number of flower visitor species (species richness) that visited alien and native species. As species richness is sensitive to sample size, we used rarefaction techniques (Gotelli & Graves 1996; Vázquez & Simberloff 2002) that allowed comparisons of species richness assuming equal number of visits. We compared the expected number of flower visitors visiting alien and native species for a sample of n = 100 visits, as most rarefaction curves had reached a clear asymptote for that sample size (Gotelli & Colwell 2001). Three native species, Lathyrus magellanicus Lam., Berberis linearifolia Phil. and Tristerix corymbosus Kuijt, were excluded from the comparison because we recorded < 100 visits. We also estimated the expected (rarefied) number of species visited by each flower visitor every 100 visits. Because most native flower visitors were recorded in only a few censuses, we restricted our comparison to the 15 native flower visitors that were recorded in more than 100 visits each. For all alien flower visitors, we recorded > 100 visits. Rarefactions were computed using the Ecosim software (Gotelli & Entsminger 2001).

We compared composition of the flower visitor assemblages associated with alien vs. native species at two taxonomic levels, order and species. To assess differences in composition of flower visitor assemblages at the higher taxonomic level, we estimated the proportion of visits made by individuals belonging to different insect orders (Hymenoptera, Diptera, Lepidoptera and Coleoptera) to alien and native species.

We characterized the structure of plant–flower visitor interactions at the species level by means of detrended correspondence analysis (DCA). We first constructed a plant × flower visitor synthetic matrix averaging all the visit frequency censuses of each flower visitor to each plant over all sites and habitat units. Each entry xij in the synthetic matrix represents the mean visit frequency per census of the flower visitor species i to the plant species j. Within the DCA, we conducted a resampling procedure to evaluate a differential association of any of the four alien flower visitors with alien plants and/or any of the five most frequent native flower visitors with native plants. For each of the most frequent alien or native flower visitors, we estimated the difference in the mean euclidean distance from each flower visitor species to the 13 alien and 15 native plant species. Each observed difference was compared with a distribution of 999 mean distance differences calculated from that flower visitor to same-sized samples chosen randomly from the list of 28 species without replacement. Euclidean distances were calculated from the scores of flower visitors and plants in the ordination diagram defined by the first and second axis of the DCA.

We further examined the relationship between the flower visitor assemblages and ecological features of the plants and their flowers, included as explanatory variables in the model, using canonical correspondence analysis (CCA; Ter Braak 1986). We carried out DCA and CCA with the canoco 4.0 package (Ter Braak & Šmilauer 1998). To date, only a few studies have used correspondence analysis to address patterns of organization in plant–pollinator communities (see Dicks et al. 2002; Potts et al. 2003), and to our knowledge only one previous study (Hingston & McQuillan 2000) has incorporated explanatory variables in order to test the statistical significance of plant features in the composition of flower visitor communities.

To characterize plant species, we selected four explanatory variables that we hypothesized to be important determinants of variation in the composition of flower visitor assemblages, including one quantitative variable (flowering time) and three qualitative variables (origin, flower symmetry and corolla colour). We constructed an index of flowering time (FT) dividing the whole flowering season, from mid-September to mid-April, into 1-month time intervals (six classes). Each species was assigned to the particular time-interval in which flowering, or the peak of flowering, took place during the sampling season. Classes of the three other qualitative variables were transformed into binary dummy variables: origin = alien or native; flower symmetry = actinomorphic or zygomorphic; and corolla colour (five classes) = red, pink, yellow, white or violet. Actinomorphic architecture included capitula of Asteraceae (see Hingston & McQuillan 2000). Appendix S1 shows the classification of each species according to these four explanatory variables.

In addition to our synthetic matrix, we also constructed a so-called site-specific matrix in which each ‘sample’ corresponds to a species sampled in a particular site, and the values in the entries correspond to the visit frequency for that particular site. Through this approach, we analysed the effect of two additional qualitative (nominal) variables on species composition: site (Puerto Blest, Llao Llao, Cerro Otto and Challhuaco) and habitat disturbance (disturbed or undisturbed). Both factors were transformed into binary dummy variables. For this analysis, we excluded the three ornithophilous species (i.e. those with red corollas) because these species were present in only one site each.

Data were log transformed [y = log10(y + 1)] to prevent a few high values from unduly influencing ordination results (Ter Braak & Šmilauer 1998). We quantified the amount of variability explained by each particular variable by means of the canoco procedure of stepwise forward selection, where the explanatory variable best fitting the data is selected first and then the next best fitting variable is added. Before each addition, the significance of the explanatory effect of the candidate variable was evaluated using Monte Carlo Permutation Test (with 999 permutations).

Results

overall patterns of flower visitor diversity and composition

We identified a total of 110 species and morphospecies among 2533 individuals visiting the flowers of 28 species throughout the 2000–01 flowering season. Of these, the 29% that were identified at least to genus level and the 21% identified to species, represented 84.6% and 83.6%, respectively, of all visits recorded. With the exception of a single hummingbird, Sephanoides sephaniodes (c. 1% of all individuals), all other flower visitors were insects belonging to the orders Hymenoptera (67%), Diptera (24%), Coleoptera (5%) and Lepidoptera (3%). We recorded four alien species visiting flowers: Bombus ruderatus Fabricius and Apis mellifera Linn. (Apidae, Hymenoptera, 382 and 119 individuals, respectively), Vespula germanica F. (Vespidae, Hymenoptera, 43 individuals), and Eristalis tenax Linn. (Syrphidae, Diptera, 27 individuals). Alien flower visitors accounted for only 3.6% of the species richness, but for 22.5% of all individuals observed on flowers.

The anova did not reveal significant differences in the mean expected (rarefied) number of flower visitor species between alien and native plant species in a random sample of 100 visits (mean ± 1 SE = 11.39 ± 1.27 vs. 10.18 ± 1.32 flower visitors, respectively; F1,23= 0.43, P = 0.518). Thus, alien plant species were not visited by a wider community of potential pollinators than were native plant species. From the point of view of the flower visitors, the expected number of plant species visited in 100 visits was not significantly different between alien and native flower visitors (8.54 ± 1.90 vs. 9.50 ± 0.98, respectively; F1,17 = 0.20, P = 0.66). Thus, native flower visitors, or at least those species that reached abundances comparable with those of alien species, were as generalized as the alien flower visitors. The ‘mutualist richness’ of both plants and flower visitors was not therefore influenced by origin.

Hymenoptera, followed by Diptera, were the most frequent flower visitors. Alien and native plant species received visits from different insect orders in a similar proportion (Table 1).

Table 1.  Results of one-way anova on the arcsine of the square root proportion of visits made by each insect order to flowers of alien and native plant species. Means and SE were back transformed
OrderProportion of visitsF(1,31)P
NativeAlien
Hymenoptera0.51 ± 0.090.50 ± 0.080.0330.856
Diptera0.36 ± 0.080.36 ± 0.030.0010.98
Coleoptera0.04 ± 0.020.07 ± 0.020.6460.429
Lepidoptera0.02 ± 0.010.03 ± 0.010.2750.604

the structure of the plant–flower visitor community

Figure 1 shows the ordination of plants and flower visitors based on the two first axes of the DCA, displaying the major variation in flower visitor composition across plant species, with plants and flower visitors on separate graphs for clarity. Most alien plant species are clearly grouped at the top left, whereas native plant species are more uniformly scattered along the first axis (Fig. 1a). Alien flower-visiting species also plot close to alien plants (Fig. 1b). Our resampling procedure showed that the four alien flower visitors were significantly closer to alien plants than to native plants (Table 2). Trends were equivocal among the most frequent native flower visitors. Scores of Bombus dahlbomii Guerin, Cadeguala albopilosa Spinola and Trichophthalma jaffueli Phil. were, on average, at a similar distance to alien and native plants. Only the native Manuelia postica Spinola was significantly closer to native than alien plant species. Contrary to expectation, however, the native Ruizantheda mutabilis Spinola was significantly closer to alien than native plants (Table 2).

Figure 1.

DCA ordination of (a) plant species and (b) flower visitor species. The two axes together explained 23.3% of the variance in visit frequency by flower visitors across plant species. The eigenvalue associated with each axis is provided in parentheses. ○ = native species; • = alien species. Plant species abbreviations follow Appendix S1. Flower visitors are coded as follows: A.m. = Apis mellifera, A.g. = Anthidium gayi, B.d. = Bombus dahlbomii, B.r. = Bombus ruderatus, C.a. = Cadeguala albopilosa, C.t. = Corynura thauca, Cor. sp. = Corynura sp., C.s. = Colletes seminitidus, E.t. = Eristalis tenax, M.g. = Manuelia gayi, M.p. = Manuelia postica, N.s. = Nothanthidium steloides, Pro. sp. = Protystes sp., R.m. = Ruizantheda mutabilis, R.n. = Ruizantheda nigrocearulea, S.s. = Sephanoides sephaniodes, T.a. = Trichophthalma amoena, T.j. = Tricophthalma jaffueli, V.g. = Vespula germanica, Cic 1 = Cicindeliadea sp.1, For 1 = Formicidae sp.1, Syr 1 = Syrphidae sp.1, Syr 2 = Syrphidae sp.2, Syr 3 = Syrphidae sp.3, Syr 4 = Syrphidae sp.4, Syr 5 = Syrphidae sp.5, Syr 6 = Syrphidae sp.6, Syr sp. = Syrphidae sp., C1 = Coleoptera 1, D8 = Diptera 8, D14 = Diptera 14, H21 = Himenoptera 21, L1 = Lepidoptera 1, L2 = Lepidoptera 2.

Table 2.  Mean Euclidean distance from scores of alien and from scores of most common native flower visitor species to scores of all alien and native plant species from the DCA ordination plot (Fig. 1) and results of the resampling procedure. Tests evaluate the null hypothesis that distance from each flower visitor score to plant scores are independent of plant species origin. (DnativeDalien) > 0 means that the score of that flower visitor species is on average closer to scores of alien than to scores of native plant species
Flower visitor speciesEuclidean distanceObservedPercentiles
DnativeDalien(DnativeDalien)0.52.597.599.5
  • Alien flower visitors,

  • *

    P < 0.05,

  • **

    P < 0.01.

Apis mellifera2.711.67 1.04**−1.13−0.79 0.750.95
Bombus ruderatus3.102.29 0.81*−1.27−0.89−0.601.19
Vespula germanica2.101.29 0.81*−0.89−0.68 0.700.96
Eristalis tenax1.401.09 0.31*−0.38−0.30 0.310.39
Bombus dahlbomii1.741.76−0.02−0.98−0.77 0.730.97
Manuelia postica2.272.98−0.71*−0.84−0.59 0.540.72
Cadeguala albopilosa1.811.35 0.46−0.86−0.64 0.690.89
Ruizantheda mutabilis2.361.52 0.84*−1.04−0.79 0.750.95
Trichophthalma jaffueli2.591.99 0.60−1.21−0.96 0.861.15

factors influencing the composition of flower visitor assemblages

Figure 2 displays the CCA ordination of plants and flower visitors. The eigenvalues of the two first axes of the DCA and of the CCA nearly coincided (Figs 1 and 2), suggesting that the chosen explanatory variables can account for a large proportion of the variation depicted in the DCA ordination (Ter Braak 1986). The CCA ordination explained 53.1% of the variance in plant–flower visitor associations.

Figure 2.

Ordination diagram of plants and flower visitors based on the two first axes of a canonical correspondence analysis (CCA) of the visit frequency of flower visitors to alien and native plant species. Plant species were classified according to four variables: origin, flower symmetry, colour and flowering time. ○ = native plants; • = alien plants; × = native flower visitors; + = alien flower visitors. Plant abbreviations follow Appendix S1. Flower visitor codes follow Fig. 1. For clarity, only plants with fit values > 5 and flower visitors with weight > 5 in the ordination are displayed. Only explanatory variables significant at P < 0.01 are shown.

The model resulting from a forward selection of explanatory variables in the synthetic matrix identified three relevant variables (in decreasing order of importance, flowering time, red corollas and species origin, Table 3). All these variables contributed significantly to the model (P < 0.01), whereas flower symmetry and all the remaining corolla colours did not contribute significantly to explain the association between plant species and flower visitors (P > 0.05).

Table 3.  Results of a forward selection of explanatory variables conducted within the CCA ordination based on (a) the synthetic matrix and (b) the site-specific matrix. Variables are listed in decreasing order of their conditional effect (i.e. their ability to explain patterns in the species data considering the effect of other explanatory variables) in the synthetic matrix. Origin = native vs. alien. Flower symmetry = zygomorphic vs. actinomorphic. Habitat disturbance = disturbed vs. undisturbed
Variable(a) Synthetic matrix(b) Site-specific matrix
λAFλAF
  • *

    P < 0.01,

  • **

    P < 0.001.

  • Additional variance each variable explains at the time it was included in the model.

Flowering time0.532.93**0.522.75**
Red corolla0.512.97*
Origin0.352.16*0.522.81**
White corolla0.241.550.140.101
Violet corolla0.221.380.150.83
Yellow corolla0.191.270.241.37
Flower symmetry0.181.160.130.71
Habitat disturbance0.341.87**
Pto. Blest0.392.18*
Llao llao0.231.33
Challhuaco0.080.46

Flowering time determined the main gradient in flower visitor species composition. The arrow representing flowering time points in the direction of maximum change in the time of flowering, i.e. flowering time is later from bottom right to top left (Fig. 2). The three species bearing red corollas were clearly separated (at the top right corner) from the other species. Thus, native species belonging to this colour class differed in their major flower visitor composition, characterized by the pre-eminence of the hummingbird Sephanoides sephaniodes. Species of different origin differed significantly in the composition of their flower visitor assemblages (Table 3, Fig. 2). The elimination of the three native species with red corollas, which are mainly ornithophilous, from the data set did not change the significance of the factor plant origin (λA = 0.36, F = 2.33, P = 0.005, where λA is the additional variance each variable explains at the time it was included in the model, see also Table 3). Thus, the influence of plant species origin persisted when only the entomophilous plant community was considered.

In general terms, species of different origin differed in flowering times (see Appendix S1). For instance, 73.3% of native and 46.2% of alien plant species bloomed during the first half of the season (time intervals 1–3), while 26.7% of natives and 53.8% of alien species bloomed during time intervals 4–6. This was reflected in a significant interaction between flowering time and species origin (λA = 0.39, F = 2.4, P = 0.002). Thus, the influence of origin on flower visitor composition differed in relation to the time of the season. Analysing early vs. late flowering species separately revealed that the influence of origin was greater late than it was early in the flowering season (λA = 0.57, F = 2.01, P = 0.01, vs. λA = 0.27, F = 1.75, P = 0.04, respectively). Accordingly, those alien species most closely associated with alien flower visitors (Carduus thoermeri, Cirsium vulgare, Hirschfeldia incana and Trifolium repens) were all late flowering (Figs 1 and 2 and Appendix S1).

Based on the site-specific matrix, the statistical significance of the relationship of the species data with flowering time and species origin was further corroborated by means of a Monte Carlo permutation test applied to a forward selection of variables in the CCA (Table 3). This approach, which evaluates the influence of each variable once the influence of variables already included in the analysis is taken into account, allowed us to separate the influence of species origin from that of habitat disturbance. As the effect of habitat disturbance could mask differences in floral composition and relative abundance of alien species, we repeated the CCA in two matrices drawn from the site-specific matrix: the first matrix only included the plant species sampled in disturbed habitats and the second matrix only included the plant species sampled in undisturbed habitats. In both habitats we found a highly significant influence of species origin on flower visitor composition (λA = 0.29, F = 1.73, P = 0.005, and λA = 0.67, F = 2.44, P = 0.005, respectively). Altogether these results demonstrated that species origin is an important determinant of flower visitor composition independent of the disturbance context that is usually associated with the invasion of alien species.

Discussion

the incidence of alien species in the native community

The flora of the temperate forests of NW Patagonia supports a diverse flower visitor community, composed mostly of native insects and a few abundant alien insects. Aliens were well integrated into the native community through plant–flower visitor interactions. However, our results show that species origin significantly influences the composition of flower visitor assemblages. Furthermore, this influence persisted after accounting for other intrinsic (flower symmetry and colour) or extrinsic (flowering time, site and disturbance) factors. In particular, alien flower visitors were more closely associated with alien than with native plant species, whereas native flower visitors were not particularly associated with plant species of any specific origin.

mutualist diversity

Alien and native plant species were associated with a similar number of flower visitor species. This result agrees with recent empirical evidence against a proposed relationship between generalization and invasibility, at least in pollination systems (see also Richardson et al. 2000; Morales & Aizen 2002), but disagrees with other studies suggesting that alien plants establish a lower number of interactions with flower visitors than native plants (Olesen et al. 2002; Memmott & Waser 2002). However, the study by Olesen et al. (2002) took place in oceanic islands where reduced interspecific competition could lead to an ecological release resulting in endemic super-generalists (Olesen et al. 2002), whereas differences from the results of Memmott & Waser (2002) could arise because our plant sample comprises only the most abundant flowering species of the entire alien flora. These authors considered a more complete flora, including rare alien plants that could have been ignored by many flower-visiting insects (Memmott & Waser 2002).

Just as for plants, we did not find any evidence that alien flower visitors are more generalized than natives. Nevertheless, it should be noted that comparisons were based on a small subset of native flower visitors (15 out of 110 species), which reached abundances comparable with those of aliens. This result agrees with the view that abundance seems to be the main determinant of the degree of pollinator specialization (Vázquez & Aizen 2003).

patterns of flower visitor composition at the community level

The flower visitor assemblages associated with alien and native species exhibited different patterns according to taxonomic level. Both alien and native plant species were visited by insects of the four main flower visitor orders (Hymenoptera, Diptera, Coleoptera and Lepidoptera, in decreasing order) in similar proportions. Yet, differences between aliens and natives emerged at the level of flower visitor species. Despite alien plant species being well integrated into the overall community-wide pollination web, Fig. 1 reveals a significantly closer association of alien flower visitors with alien plants than with native plants. Thus, those alien flower visitors, which in spite of representing a small proportion of all species accounted for > 20% of total individuals recorded visiting flowers, might be in part driving the higher similarity in flower-visiting assemblages among alien plants. As the most abundant native flower visitors were as closely associated with alien as with native species, alien plant species and their associated flower visitors may be viewed as a subnetwork embedded within the overall plant–flower visitor network.

As almost all alien plant species, except Lupinus polyphyllus, arrived from Europe, European alien flower visitors may have integrated their flower visitor assemblages in their native ranges. If so, the closer association between alien plants and flower visitors could arise from past coevolution in their ancestral region (Richardson et al. 2000). An alternative explanation is that those characteristics of plants and insects that confer the capacity to spread and invade, might also promote a closer match between them (Morales & Aizen 2002). For instance, Apis mellifera was the species most closely associated with alien plants. In a previous study, a higher proportion of visits made by alien flower visitors to alien plants, in comparison with native plants, was explained by a strong preference of A. mellifera for alien species (Morales & Aizen 2002). Most alien species here produce large and showy inflorescences (e.g. Taraxacum officinale, C. vulgare, C. thoermeri, L. polyphyllus and T. repens) or present high flower density per area (e.g. C. scoparius, H. incana) (C. L. Morales, unpublished results). Apis mellifera is able to exploit high resource patches much more effectively than their solitary counterparts by virtue of its foraging system (Potts et al. 2003). In addition, alien social bees, like A. mellifera and Bombus ruderatus, need high energy acquisition for colony maintenance (Heinrich 1979), which may make it profitable to exploit the highly concentrated and abundant nectar and pollen resources per flower, inflorescence or area provided by some alien plant species (Chittka & Schurkens 2001; Schurkens & Chittka 2001; Brown et al. 2002). The formation of dense flowering patches through massive local recruitment is a phenomenon often associated with plant invasions (Schurkens & Chittka 2001; Ghazoul 2002).

Our results support the existence of ‘invader complexes of mutualists’, defined as groups of introduced species interacting more with each other than expected by chance (sensu Olesen et al. 2002). Focusing on the plant–flower visitor webs of two different oceanic islands, Olesen et al. (2002) did not find any evidence for such invader complexes. However, the existence of many endemic ‘supergeneralists’ in those islands (see above), which promptly incorporate alien species as mutualists, might hinder such alien/alien associations.

other factors influencing flower visitor composition

Species of different origin differed in their flower visitor assemblages independent of any other studied factor. Beyond species origin, flowering time, bearing red corollas and, to a lesser extent, habitat disturbance also explained a significant part of the variation in the flower visitor assemblages.

Despite a relatively short flowering season typical of temperate latitudes, a temporal turnover of flower visitor species characterized the flowering community. This turnover led to major differences in the flower visitor assemblages associated with species flowering at different times over the season. In other studies, flowering phenology had a larger effect on visitor profiles than visible floral traits (Herrera 1988; Hingston & McQuillan 1999), or influenced how a plant–pollinator web breaks up into compartments (Dicks et al. 2002). Phenological matching seems to be a fundamental factor in plant–animal interactions that is not exclusive to pollination. For instance, phenological uncoupling accounted for > 50% of the unrecorded links in a plant–animal seed dispersal web (Jordano et al. 2003). As the composition of flower visitor assemblages associated with both alien and native plants was subjected to this phenological constraint, the impact of alien plant species on the pollination of native species might depend to a large extent on the degree of phenological overlap.

The influence of red corollas on flower visitor composition is related to a separation of typically ornithophilous from entomophilous species (Fig. 2), even when the former also received visits by insects as well as visits by Sephanoides sephaniodes. No alien species was visited by hummingbirds, and no typical ornithophilous alien species has invaded these forests so far. The dependence on a single hummingbird species for pollination might reduce the chance of successful invasion of ornithophilous species, once the organism has reached the new location. Thus, the (direct) interaction between native hummingbirds and the alien flora can be considered negligible in our system.

Habitat disturbance per se also influenced the composition of plant flower visitor assemblages. In a previous work, we found that the proportion of visits by B. ruderatus was significantly higher in disturbed habitats, regardless of species origin (Morales & Aizen 2002). This agrees with the results of Aizen & Feinsinger (2003), who found that the relative abundance of B. ruderatus gradually increased along a disturbance gradient. In addition, insect sampling using water traps at the same study sites showed that some native species are more abundant in disturbed habitats (C. L. Morales et al., unpublished). Thus, flower visitors seem to differ in sensitivity to habitat disturbance.

The lack of influence of flower symmetry and the remaining flower colours on flower visitor composition suggests the absence of floral syndromes (i.e. flower adaptations to particular pollinator groups) related to different insect orders in our entomophilous plant community (see Faegri & Van der Pijl 1979). In addition, the array of flower functional types found among alien species (despite the absence of ornithophilous species) does not seem to differ strongly from that among natives, which could explain the similar relative flower visitor composition at the level of insect orders.

Overall, our results show that differences in flower visitor assemblages between alien and native species could not be explained by any other significant factor influencing the structure of the studied plant–pollinator web (i.e. flowering time, flower symmetry and colour or habitat disturbance). However, we still need to explore what other potential covariates of species origin might help to explain why species of the same origin are more similar to each other in terms of pollinator assemblages. Other traits could include type, quantity and quality of floral reward, density of individual plants and flowers, and taxonomic affinity.

conservation implications

A preferential association between alien plants and flower visitors could have deep conservation implications (Richardson et al. 2000; Barthell et al. 2001). Overall, our results support the hypothesis of biotic facilitation (sensuSimberloff & von Holle 1999), which states that, upon arrival in a new region, alien species establish not only antagonistic but also mutualistic interactions (e.g. pollination) with the species already present (Olesen et al. 2002). In particular, we demonstrated the existence of invader complexes that may synergically favour alien partners through pollination (see also Morales & Aizen 2002). This phenomenon might be exacerbated in disturbed habitats, where aliens are more abundant. On the other hand, the influence of alien plants on pollination of native plants seems to be limited to those plants that overlap in blooming. As some of the alien plant species reach high densities and have conspicuous flowers, particular attention should be paid to those native species with which they overlap in flowering time, in order to detect possible changes in visitation rates and seed output.

Acknowledgements

We thank A.M. Klein, D. Vázquez, L. Galetto, A. Lack and an anonymous reviewer for valuable comments on an earlier manuscript. J. Betinelli and J. Lichtstein helped during fieldwork. We are indebted to Arturo Roig for help in identifying bees. We thank also C. Ezcurra and J. Puntieri for plant identification. Funding was provided by Canon Inc. USA, the AAAS and the USNPS through the Canon National Parks Science Scholars Program for the Americas. C.L.M. holds a doctoral scholarship of the National Research Council of Argentina (CONICET). M.A.A. is a member of the Carrera del Investigador from the same institution.

Ancillary