Robert R. Junker, Department of Animal Ecology and Tropical Ecology, University of Würzburg, Biozentrum, Am Hubland 97074, Würzburg, Germany. E-mail: firstname.lastname@example.org
1. Sympatric flower visitor species often partition nectar and pollen and thus affect each other's foraging pattern. Consequently, their pollination service may also be influenced by the presence of other flower visiting species. Ants are solely interested in nectar and frequent flower visitors of some plant species but usually provide no pollination service. Obligate flower visitors such as bees depend on both nectar and pollen and are often more effective pollinators.
2. In Hawaii, we studied the complex interactions between flowers of the endemic tree Metrosideros polymorpha (Myrtaceae) and both, endemic and introduced flower-visiting insects. The former main-pollinators of M. polymorpha were birds, which, however, became rare. We evaluated the pollinator effectiveness of endemic and invasive bees and whether it is affected by the type of resource collected and the presence of ants on flowers.
3. Ants were dominant nectar-consumers that mostly depleted the nectar of visited inflorescences. Accordingly, the visitation frequency, duration, and consequently the pollinator effectiveness of nectar-foraging honeybees (Apis mellifera) strongly decreased on ant-visited flowers, whereas pollen-collecting bees remained largely unaffected by ants. Overall, endemic bees (Hylaeus spp.) were ineffective pollinators.
4. The average net effect of ants on pollination of M. polymorpha was neutral, corresponding to a similar fruit set of ant-visited and ant-free inflorescences.
5. Our results suggest that invasive social hymenopterans that often have negative impacts on the Hawaiian flora and fauna may occasionally provide neutral (ants) or even beneficial net effects (honeybees), especially in the absence of native birds.
Resource partitioning in response to inter-specific competition stabilises the coexistence of species that occupy similar niches (Arthur, 1986). Interactions between sympatric flower visitors have been frequently studied and it has been shown that the presence or absence of one species affects the resource utilisation of other species, either influencing the plant species visited (Inouye, 1978; Nagamitsu & Inoue, 1997) or the temporal (Stone et al., 1996) and spatial (Morse, 1982) pattern of visitation to individual plants. Many alien flower-visiting species (e.g. Apis mellifera) are particularly dominant competitors (Traveset & Richardson, 2006), displace native pollinators (Kato et al., 1999) and thereby potentially have negative effects on the reproduction of native plants (Traveset & Richardson, 2006) when their pollinator effectiveness is lower than the effectiveness of the native, displaced pollinators (see Madjidian et al., 2008), but note their importance as pollinators in many introduced areas (Dick, 2001; Goulson, 2003).
Several studies demonstrated that pollinator effectiveness differs between different flower visitors of the same plant species (Kandori, 2002; Madjidian et al., 2008; Reynolds & Fenster, 2008). Only a few studies have examined how nectar or pollen collection potentially affects effectiveness (Free, 1966; McIntosh, 2005). The impact of other visitors on the foraging behaviour of a flower visitor and consequences for effectiveness has not been tested so far. Pollinator effectiveness is the product of a floral visitor's quantity and quality of actions leading to pollination (Herrera, 1989; Kandori, 2002; Reynolds & Fenster, 2008). The visitation rate, the duration of visits, the frequency of contacts to stigmas, and the amount of pollen deposited on stigmas per contact are surrogates to evaluate effectiveness. The fruit or seed set after visits of individual species represent more direct measures of effectiveness.
Ants and bees visit flowers in search of a different set of resources: ants as facultative flower visitors (Junker & Blüthgen, 2010) do not depend on floral resources and are solely interested in sugar-rich but usually amino acid-poor nectar. Bees as obligate flower visitors (Junker & Blüthgen, 2010) collect both nectar and amino acid-rich pollen, which cannot be digested by most ants with a few exceptions (Urbani & deAndrade, 1997; Blüthgen & Feldhaar, 2010). Thus, regarding pollen, ants and bees segregate the resource. The consumption of nectar, however, may be influenced by inter-specific competition. These contrasting interests may lead to complex interactions between ants and bees, particularly on flowers where both resources (nectar and pollen) are spatially separated within one flower.
On the Hawaiian Islands where no ants were present prior to their human introduction, we studied the interactions between ants and bees on flowers of the endemic Hawaiian tree Metrosideros polymorpha Gaudich. (Myrtaceae). M. polymorpha has brush-like flowers where nectar and pollen are spatially separated. Thus, these flowers are well adapted to bird pollination (Carpenter, 1976). Nowadays, the flowers are numerously visited by social hymenopterans such as ants and bees (Lach, 2005, 2008a). Lach (2005) observed that invasive ants deplete and defend the nectar of visited flowers and thereby reduce the visitation frequency of endemic Hylaeus spp. bees (2008a). In the present study, we addressed three questions. (i) How do invasive ants and honeybees (Apis mellifera L.) and endemic Hylaeus bees partition the resources provided by the flowers of M. polymorpha? (ii) What are the behavioural responses by introduced and endemic bees to the presence of ants? Do these responses depend on the resources collected by the bees? (iii) How does the bees' pollinator effectiveness change with the type of resource collected and the presence or absence of ants?
Materials and methods
Study sites and organisms
This study was performed on the island of Hawai'i in Hawai'i Volcanoes National Park (HAVO) from March to May 2009. We selected two sites within HAVO where M. polymorpha (Myrtaceae) trees dominated the vegetation and provided the most floral resources within the area. Alanui Kahiko (188 m a.s.l.; N 19°18.18′, W 155°08.93′) is an old lava flow with sparse ground-vegetation and was invaded by Anoplolepis gracilipes F. Smith ants. The second site, the Broomsedge Burn Area (1230 m a.s.l.; N 19°26.23′, W 155°17.97′), was densely covered with the naturalised grasses Andropogon virginicus and Schizachyrium condensatum (Poaceae) and was inhabited by Linepithema humile Mayr ants.
Metrosideros polymorpha is found from sea level to tree line (0–2600 m a.s.l.) in a variety of habitats and successional stages and occurs in various growth forms from small shrubs to large trees (Wagner et al., 1990). The brush-like inflorescences are composed of 10–30 flowers that open sequentially either within a long period (up to 40 days) or a short time span (Carpenter, 1976). The open flowers are cup-like with a central style and numerous filaments encircling the nectar cup. The length of style and filaments was up to 3.5 cm in our study trees but may be much shorter in other populations (Wagner et al., 1990). The proportional fruit set by the partially self-incompatible flowers is usually much higher after pollination (Carpenter, 1976), but see Burton (1982). The flowers are regarded as bird pollinated (Carpenter, 1976) and were historically visited by native honeycreepers (Drepanididae), but today also by introduced bird species (e.g. Zosterops japonicus) (Carpenter, 1976). Only occasionally (approximately once a week during morning hours) we observed birds (e.g. Himatione sanguinea) foraging on M. polymorpha flowers at both sites. Several species of hymenopterans also frequently visit the flowers and some of them also contribute to pollination (Carpenter, 1976; Lach, 2008a). The flowers in both study sites where visited by invasive honeybees A. mellifera, and the flowers in Alanui Kahiko were also visited by endemic bee species of the genus Hylaeus. These bees are much smaller and less abundant than honeybees. The flowers in Alanui Kahiko where visited by A. gracilipes ants (body size ∼5 mm), those in the Broomsedge Burn Area by Linepithema humile ants (body size ∼2.5 mm).
Treatment of inflorescences
In each of the study sites, 10 trees were selected that featured sufficient flower buds. Several inflorescences on each tree were treated in one of the following ways. (1) Ant exclusion: a large branch, often bearing several inflorescences, was covered with a sticky barrier (Raupenleim, Schacht, Germany) at its base, and ants currently visiting the branches were removed. Thus, only flying animals had access to the flowers. (2) Control: untreated branches – inflorescences were accessible to all visitors. (3) Ants only: one inflorescence per tree was haphazardly selected and encased with a plastic cup (250 ml) with the bottom facing the base of the branch. We cut a hole (one-third of the circumference) in the bottom to allow ants to crawl in and a second notch for the branch. The top was covered with a pollen-tight webbing to exclude flying insects and wind-drifted pollen. A strip of the plastic cup was cut and likewise covered with pollen-tight mesh in order to reduce the ‘green-house effect' inside the cups. We used cups instead of mesh bags in order to prevent the mesh touching the reproductive structures and thereby contributing to pollination. (4) Complete exclusion: one inflorescence of the ant exclusion branch (1) was haphazardly selected and encased with a cut plastic cup similar to (3). Places of treatments were chosen at a similar height within each tree and facing towards the same compass direction in order to reduce variability caused by abiotic parameters.
Between 06.30 and 12.00 hours, from a distance of 1–2 m we observed all ant-visited and ant-free inflorescences (treatments 1, ant exclusion and 2, control) of a tree and noted the flower visitor species (i.e. Apis mellifera or Hylaeus spp. that were the only frequent visitors next to ants), duration of visits, presence/absence of ants, resource used by flower visitor (pollen or nectar), and the total observation time. At the end of each observation period the number of ants inflorescence−1 and flowers inflorescence−1 was noted.
On 8 days between 07.30 and 14.00 hours the standing crop of nectar (µl) in flowers of ant-visited (n = 92) and ant-free branches (n = 92) was determined using micro-capillaries (5 µl). Sugar concentration was measured with a handheld refractometer (Eclipse, Bellingham and Stanley, Tunbridge Wells, U.K.).
In another test series, we carefully approached the inflorescences after bees had landed on ant-visited branches (treatment 2, control) and observed interactions between bees and ants (n = 98). Interactions were assigned to aggressive interactions (bees left the inflorescences after contact or ant approaching) and non-aggressive interactions (ants and bees shared the inflorescence).
Stigma contacts and pollen deposition
The number of stigma contacts during flower visits (contacts min−1) of A. mellifera (n = 48visits) and Hylaeus spp. (n = 26) was counted and it was noted which resource the bees collected. Furthermore, 27 inflorescences were enveloped with a wire frame covered with pollen-tight webbing prior to anthesis in order to prevent contamination of stigmas with pollen by animals or wind. During anthesis the wire frame was removed and three stigmas were immediately removed using forceps and were prepared for microscopy. Stigmas were placed between an object slide and a cover slip and were gently squeezed. Afterwards, inflorescences with the remaining stigmas were free to be visited by bees and stigmas were also prepared for microscopy after the first contact. The preparations were stained with basic fuchsine-solution (see Kearns & Inouye, 1993) and pollen grains were counted approximately 10 min after staining, when the staining caused a pink colouration of the stigma-tissue while the pollen grains not yet absorbed the staining and remained yellow.
Pollinator effectiveness was estimated for nectar and pollen collecting honeybees and Hylaeus bees on ant-free and ant-visited flowers. We defined the pollinator effectiveness E (pollen h−1) as product of the visitation frequency f (visits h−1), the visitation time t (h visit−1), the stigma contacts c (contacts h−1), and the pollen deposition d (pollen contact−1). Several studies demonstrated that the amount of pollen deposited decreases over subsequently visited flowers (Thomson & Plowright, 1980; Price & Waser, 1982; Waser & Price, 1982; Johnson & Nilsson, 1999; Adler & Irwin, 2006). Therefore, it is unlikely that the amount of pollen deposited on stigmas within the same inflorescence linearly increases with subsequent contacts to the stigmas. To account for this saturating function, we introduced a power term (≤ 1) to the product of t and c, which thus diminishes the pollen quantity deposited during subsequent stigma contacts. Using the following formula for pollinator effectiveness E, we assumed that the first stigma contact by a bee leads to the measured deposition of pollen (see above) while the following stigmas receive a smaller proportion of the initial number.
Effectiveness was calculated for power-terms ranging from 0.1 to 1 with stepwise increments of 0.1. We used bootstrapping (1000 replications) to calculate the mean pollinator effectiveness and its confidence interval for both bee taxa as a function of resource collected and presence/absence of ants. A combination of the specific data-sets f, t, c, and d for nectar- and pollen-collecting honeybees and Hylaeus spp. bees on flowers with and without ants were separately resampled (with replacement) for n = min(nf,nt,nc,nd) times and we calculated the mean of their products.
On all treated trees, the number of flower buds and flowers of three inflorescences of treatments 1 and 2 (ant exclusion and control) and the inflorescences of 3 and 4 (ants only and complete exclusion) were counted before or during their anthesis and compared with the number of developing fruits a few weeks later (total n = 57). As a result of a partial self-incompatibility of red flower morphs, which were present in our study sites, maximal fruit set occurs only after cross-pollination (Carpenter, 1976).
Linepithema humile ants visited the flowers (treatment 2, control) of M. polymorpha in the Broomsedge Burn Area in high densities (mean ±SE: 1.0±0.16ants flower−1), whereas A. gracilipes ants visited the flowers in Alanui Kahiko in lower densities (0.13 ± 0.02) at a given time. Honeybees (Apis mellifera) collected nectar and pollen of M. polymorpha flowers at both sites, whereas we only observed Hylaeus spp. bees in Alanui Kahiko, collecting pollen. Overall, ant-free inflorescences were more frequently visited by bees (mean ±SE: 0.25 ± 0.09 bees flower−1 h−1) than ant-visited ones (0.07 ± 0.02) on the same tree individual, albeit this was only marginally significant (paired t-test : t25 = 2.01, P = 0.055). With an average of 15.6 ± 1.4 (mean ±SE) flowers per inflorescence, the visitation rate amounts to 3.9 bees per ant-free inflorescence per hour. Nectar-foraging honeybees strongly preferred ant-free flowers over ant-visited ones (paired t-test: t14 = 2.2, P = 0.049; Fig. 1), whereas pollen-collecting honeybees and Hylaeus did not discriminate between the flowers (t≤ 0.79, P≥ 0.45) (Fig. 1). The same is true for the time spent on inflorescences: pollen-collecting honeybees and Hylaeus remained similarly long on inflorescences with and without ants (Welch corrected t-test: t≤ 0.503, p≥ 0.62) (Fig. 1). Nectar-collecting honeybees, however, remained twice as long on ant-free inflorescences than on ant-visited ones (t44.1 = 3.3, P < 0.01, Fig. 1).
The proportion of nectar to pollen-collecting honeybees (proportion = nectar collectors/pollen collectors) was typically lower on ant-visited (1.00 ± 0.02) inflorescences than on ant-free ones (1.25 ± 0.11). Thus, the relative frequency of nectar foragers significantly increased on ant-free inflorescences, whereas the relative frequency of pollen foragers decreased compared with ant-visited inflorescences and vice versa (paired t-test : t20 = 2.48, P = 0.022). This resource shift was independent of the density of ants visiting the flowers for a data set where results from both areas with different ant species were pooled (Pearson's R2 = 0.08, d.f. = 19, P = 0.23). The resource shift was also independent of the density of L. humile ants in the Broomsedge Burn Area alone (Pearson's R2 = 0.12, df = 6, p = 0.41), but it was negatively correlated with the density of A. gracilipes ants in Alanui Kahiko (Pearson's R2 = 0.53, d.f. = 11, P < 0.01), which was, however, the result of a single outlier.
Ants dramatically reduced available nectar: on average, flowers of inflorescences where ants were excluded (treatment 1, ant exclusion) provided 11 times more nectar (mean ± SE: 6.11 ± 0.97 µl flower−1) than flowers where ants had access (treatment 2, control, 0.55 ± 0.35 µl flower−1) (paired t-test: t22 = 5.3, P < 0.001). This effect was equally pronounced in both sites (t≥ 2.89, P≤ 0.023). Average sugar concentration of nectar was 32.1 ± 2.6%w/w (mean ± SE, n = 45).
In total, we observed 98 cases where ants and bees shared an inflorescence; in only 13.3% of all cases ants displayed aggressive behaviour and displaced bees. Bees and ants often collected resources, respectively, on the same or an adjacent flower without an interaction. However, most of the observations (91.8%) involved pollen-collecting bees as nectar-foragers on ant-visited flowers were rare. One-third of interactions that involved nectar-collecting bees were aggressive, but note the small sample size (n = 6).
Stigma contacts and pollen deposition
Number of stigma contacts was highly dependent on the bee species and the resource the bees collected (anova: F2,71 = 88.6, P < 0.001) (Fig. 2): nectar-collecting honeybees and pollen-collecting Hylaeus rarely touched any stigmas, whereas pollen-collecting honeybees had a high frequency of stigma contacts. Furthermore, honeybees deposited significantly more pollen per stigma contact on the receptive structure than Hylaeus. After contacts by Hylaeus, stigmas did not contain more pollen grains than control stigmas (F2,44 = 5.6, P < 0.01, Fig. 3). Ants very rarely climbed up the styles or filaments and thus almost never had contact with the stigmas and thus most probably do not contribute to pollination.
In the following, the results of pollinator effectiveness E calculated with the power-term = 0.5 are described. However, ranking of E (as in Fig. 4) was largely independent (see Appendix S1). The presence/absence of ants and the resource collected (nectar or pollen) by bees strongly influenced the effectiveness E of the pollinators (anova : F5,5994 = 368.3, P < 0.001, Fig. 4). Honeybees that collected nectar were much more effective on ant-free flowers than on ant-visited flowers (Fig. 4) – a result caused by a reduced visitation frequency f and time t spent on flowers (Fig. 1). Pollen-collecting honeybees on ant-visited flowers, however, were more effective than those on ant-free flowers (Fig. 4). This results from a slightly (but not significantly) higher visitation frequency of pollen-collecting honeybees on ant-visited flowers. Effectiveness E of pollen-collecting Hylaeus was not affected by the presence of ants in inflorescences (Fig. 4). Overall, the effectiveness of both bee taxa, regardless of whether nectar or pollen was collected, differed slightly, but significantly, between ant-visited (mean ± SE: E = 9.1 ± 0.26) and ant-free flowers (12.5 ± 0.37) (t5350.7 = 7.45, P < 0.001). The average pollinator effectiveness of honeybees (14.5 ± 0.31) was much higher than that of Hylaeus (3.4 ± 0.19) (t5933.8 = 30.5, P < 0.001).
Fruit set strongly varied between treatments (anova: F3,116 = 22.8, P < 0.001) (Fig. 5): it was highest in inflorescences to which all visitors had access (control, 2) and where ants were excluded but flying insects were allowed to visit (treatment 1, ant exclusion). Inflorescences from which all visitors were excluded (treatment 4, complete exclusion) and inflorescences where only ants had access (treatment 3, ants only) had the lowest fruit set.
The present study revealed novel insights into the interactions between flowers and different flower visiting insect species and how resource partitioning between the species affects pollinator effectiveness.
The presence of ants strongly affected the resource utilisation of bees. Ants were competitively dominant nectar-foragers on M. polymorpha, where they almost completely depleted the nectar of flowers and sometimes (in less than 15% of all observed encounters) actively defended this resource against bees. Thus, exploitation and, to a lesser extent, interference competition led to intra-floral resource partitioning: the frequency of nectar foraging honeybees strongly increased on ant-free flowers, while ants did not strongly affect the frequency of pollen-collectors. This dichotomy can be explained by the ants' interest in nectar but not pollen, and by the floral morphology of M. polymorpha where both resources are spatially separated from each other. Thus, ants stayed only in the cup-like structures that bear the nectaries and where nectar accumulates, whereas the anthers and the stigmas remain disregarded by the ants. The absence of nectar-collecting Hylaeus bees remains unexplained as they are known to collect nectar from other plant species (Magnacca, 2007). Our results partly support results from Lach (2005, 2008a) who also observed the effects of A. gracilipes and L. humile ants on bee visitation on M. polymorpha flowers. However, Hylaeus bees were shown to strongly reduce their visitation frequency on flowers visited by Pheidole megacephala (Lach, 2008a) although they also collected pollen only (L. Lach, pers. comm.). In Lach's study (2008a), honeybees were not affected by ant-visits in terms of their visitation frequency or length of visitation but no information on the resource collected by honeybees is given.
Furthermore, while Lach (2005) found that A. gracilipes strongly outcompetes bees by interference and exploitation, and L. humile mainly by interference, our data suggest a strong exploitation and weak interference competition for both species. The extent to which ants deter bees from flowers may be dependent on the ant-species present mainly because of different aggression levels (Ness, 2006). The two ant species observed in the present study occupied flowers in different densities, thus we were not able to separate effects of density and species. Furthermore, as the main effect of ants was the exploitation of the nectar reward where both ant species succeeded equally well, no ant species-effects may be expected in this case.
The identity of the bees, the ants' presence or absence, and the utilisation of either nectar or pollen, led to a pronounced variability of the pollinator effectiveness. The smaller Hylaeus bees rarely touched stigmas whereas the larger honeybees inevitably had multiple contacts with stigmas while collecting pollen. Flowers of M. polymorpha are typically bird pollinated (Carpenter, 1976) and thus the flowers' morphology is not adapted to the Hylaeus bees' small body size. The structural mismatch between flower visitors and flowers often leads to pollen theft (Hargreaves et al., 2009), which may be pronounced in this specific interaction. Invasive, feral honeybees are assumed to have multiple negative effects on native ecosystems (Goulson, 2003) but are also known as long-distance dispersers of pollen in native plants (Dick, 2001). In the present study, honeybees were effective pollinators of M. polymorpha in terms of pollen deposition per time, whereas endemic Hylaeus bees deposited very few pollen on stigmas and may thus not be regarded as pollinators. Therefore, as bird populations have declined in Hawaii (Benning et al., 2002), the presence of introduced honeybees may ensure the pollination of this endemic species.
Although ants are effective pollinators for some plant species (Gomez & Zamora, 1992), they do not contribute to pollination in M. polymorpha as they rarely contact anthers and stigmas of these flowers and the fruit set of inflorescences that were visited by ants only was low. The ants' effect on overall pollination effectiveness by the bee taxa may be ignored. On average, ant-visited flowers were only slightly less effectively pollinated than ant-free flowers. This result was reflected in the fruit set of flowers from which ants have been experimentally excluded over the whole period from flower maturation to fruit set, but to which flying insects had access. The high proportion of fruit set in bee-visited flowers suggests that bees deposited not only pollen from the same but also from other individuals. We seldom observed birds visiting flowers of M. polymorpha during that time of the day we spend in the field and the strong decline in the bird populations in recent years (Benning et al., 2002) suggest a low contribution of birds to the pollination of our study trees.
Madjidian et al. (2008) identified some restrictions regarding estimates on pollinator effectiveness that also apply to our study. The most important restrictions in our system may be: (i) our measurements of parameters incorporated in the pollinator effectiveness were ‘snapshots’ and it is unclear whether visitation frequencies and times remain constant over the whole receptive period of flowers. (ii) The estimate of the effectiveness is only meaningful if pollen is limited, which is likely for M. polymorpha as a high percentage of the seeds are non-viable (Drake, 1992). (iii) The effectiveness of Hylaeus bees may be overestimated as the number of pollen grains deposited on stigmas was not significantly higher than on control stigmas. (iv) It is unclear whether duration of visit is positively or negatively correlated with pollinator effectiveness. We assumed a positive relationship, which is supported by some studies (Ivey et al., 2003) but the opposite is suggested in others (Gomez & Zamora, 1999). As Hylaeus bees spent comparatively long times on individual flowers or anthers, the proportion of pollen deposited in stigmas from other tree individuals may decrease over the time and consequently the contribution to pollination as well. The introduction of the power term to the estimated pollinator effectiveness strongly reduces the benefit from a longer visitation duration on an inflorescence which may provide a more realistic estimation than without the power term. But note that the relative contributions of different bees as a function of resource collected and the impact of ants remain largely unaffected by the choice of the power term.
The present study demonstrated that the pollinator effectiveness of flower visitor species is highly variable and can be altered by intra-floral resource partitioning between endemic and invasive flower visitors. We did not find any direct negative effects of invasive ants and bees on endemic bees; but negative effects resulting from competition for nesting sites or resources and predation are demonstrated in other studies (Gross, 2001; Magnacca, 2007). These results suggest that invasive social hymenopterans that have devastating effects on the native Hawaiian flora and fauna (Medeiros et al., 1986; Holway et al., 2002; Goulson, 2003; Krushelnycky et al., 2005; Krushelnycky & Gillespie, 2008; Lach, 2008a) may also have neutral (ants) or even positive effects (honeybees) on endemic plant species.
We thank Rhonda Loh for her support in HAVO and Lori Lach for valuable comments on the manuscript, Diana Tichy for statistical help and the United States Department of the Interior (National Park Service) for permits to work in HAVO. The project was financially supported by the Deutsche Forschungsgemeinschaft (DFG, BL 960/1-1), the Evangelisches Studienwerk e.V. Villigst, and by the Deutscher Akademischer Austauschdienst with a travel support for R.B.