Introduction

Pollinators can drive positive indirect interactions when comingled plant species attract more pollinator species and visits than the same plant species would attract growing alone (Feldman, 2004; Lazaró et al., 2009; Braun and Lortie, 2019; Underwood et al., 2020). Indirect interactions among plants involving pollinators can also be competitive (Levin, 1970; Reader, 1975; Waser and Fugate, 1986; Underwood et al., 2020), with interspecific neighbors either reducing visitation in species-diverse mixtures or reducing “carry over,” the amount of pollen reaching a second conspecific after a pollinator stops over at a different species. Such positive and negative indirect interactions are common because many pollinators are generalists (Goulson, 2003; Hingston and McQuillan, 1998) and prefer mixed diets. Furthermore, generalist pollinators are also disproportionately abundant relative to specialists (Fort et al., 2016; Vázquez and Aizen, 2003), which is important because their abundance can amplify their effects. For example, nearly a hundred plant species can provide floral resources to honey bee (Apis mellifera) colonies (Coffey and Breen, 1997), and globally, honey bees utilize nearly 40,000 different species (Crane, 1990). Hingston and McQuillan (1998) found that Bombus species in Tasmania visited over 66 plant species from 21 different plant families, and Macfarlane (1974) found that Bombus species visited 419 different plant species in New Zealand. In Brazil, Filho and Packer (2015) found that the bees in the Megachilidae family visited 112 different plant species.

Most work on indirect facilitation mediated by pollinators has focused on “magnet species” (Braun and Lortie, 2019). Magnet species are those in pollination webs that attract a highly disproportionately large number of pollinators and enhance the pollination success of their less attractive neighbors because their unusual attractiveness (i.e., Walther et al., 2008). This is due to flexible foraging strategies, in which insect pollinators switch from magnet species to other nearby flowering species. These switches may be influenced by patchy floral distribution, depletion of floral resources by other foragers, and low encounter rates of the magnet species (Goulson,1999, 1997). For example, Podophyllum peltatum (mayapple) produces no nectar, and Laverty and Plowright (1988) found that Podophyllum plants that were near Pedicularis canadensis (lousewort) produced more fruits and seeds than plants that were far from Pedicularis. This suggests that Pedicularis acted as a magnet for pollinators because of its showy flowers and nectar. Johnson et al. (2000) transplanted the non-nectar-producing orchid, Anacamptis morio, into patches of nectar-producing plants and found that visits to the orchid increased in the presence of the nectar-producing species. Master and Emery (2015) found a 240% increase in pollinator visitation to plots that contained the magnet Ranunculus ficaria compared to plots without Ranunculus. Clearly, exceptionally attractive plant species can indirectly facilitate other plant species that are pollinated by insects.

Magnet species may drive indirect facilitation, but focus on unusually attractive species might overestimate the generality of such positive effects in communities (Braun and Lortie (2019). In other words, might such indirect facilitative interactions function without magnet species? Can such indirect interactions be caused by mixtures of species that are similar in attractiveness? There is some evidence that such diversity-based facilitation may occur (see Braun and Lortie, 2019). Ghazoul (2006) found that pollinator visits to Raphanus raphanistrum, and subsequent seed production, increased when it co-occurred with Hypericum perforatum or Solidago canadensis instead of when Raphanus was alone. This suggests that interspecific facilitation involving pollinators may not require magnet species, but the attractiveness of Hypericum and Solidago, relative to Raphanus, was not measured. Lazaró et al. (2009) found that the diversity of plant species surrounding six different target species correlated with the composition of the pollinator community that visited the target, with visitation by uncommon pollinator species increasing with surrounding floral diversity. Thus, there is a growing, but limited, body of evidence, suggesting that indirect facilitation mediated by pollinators may be common even without the unusually strong effects of magnet species (Braun and Lortie, 2019).

Other plant community attributes, such as plant species density, can also have strong effects on pollinator visitation, and these can be stronger than the effects of species richness (Borges et al., 2003; Dauber et al., 2010; Hegland and Boeke, 2006; Sih and Baltus, 1987; Munoz and Cavieres, 2008) and alter the effects of species richness (Braun and Lortie, 2019). For example, Jennersten (1988) compared the pollination success of Dianthus deltoides in small fragmented populations to that in larger continuous populations. Dianthus flowers received fewer visits from pollinators in small populations than in large populations and seed set was much lower. In small populations experimentally pollinated flowers substantially increased seed set, but this did not occur in large populations. Platt et al. (1974) experimentally manipulated population densities of Astragalus canadensis and found that seed production was higher for high-density populations than low-density populations, and attributed this to pollination success and host detection by the pollinators. Host density clearly affects pollinator visitation, but we know little about the relative importance of host population density versus host community richness.

In order to explore the effects of floral richness in the context of the effects of flowering plant density and different degrees of inherent attractiveness, we used small-scale experimental arrangements of native intermountain prairie species to explore how variation in species richness and density affected pollinator visitation rates, diversity, and behavior. We asked the following questions: (1) Do flowering species attract more pollinator visits when planted with other flowering species? (2) Do two-species mixtures attract more visits, more diverse pollinator species, and longer residence times than single species? And, (3) do two-species mixtures produce higher seed productivity per mature inflorescence (seed head) than single species?

Methods

Study system

Our study was conducted in a common garden at Fort Missoula, Missoula Montana (N 46.84278 and W-114.05806) during May and June 2015 (see Liao et al., 2015). We used 10 perennial species that commonly occur in grasslands in the northern Rocky Mountains to create native plant assemblages. These were five grasses (Festuca idahoensis, Koeleria macrantha, Poa secunda, Pseudoroegneria spicata, and Stipa comata) and five forbs (Achillea millefolium, Artemisia frigida, Gaillardia aristata, Geranium viscosisimum, and Linum lewisii). The grasses did not provide resources to pollinators but their presence provided a realistic environment for pollinators to forage in. The five forb species increased our chances of getting more than one species flowering at the same time in plots. In February 2013, we grew plants from field-collected seeds in 50-ml pots filled with 50% sand and 50% potting soil in a greenhouse at University of Montana, Missoula, USA. Three months later, seedlings were transplanted into randomly arranged 0.8 × 1.0 m2 plots in the common garden, which would be defined as “small-grained” by Braun and Lortie (2019). The distance between plots was 30 cm. We established monocultures for each individual species (n = 4 per species) and polycultures with all 10 species (n = 10) for a total of 50 plots. In each monoculture, twenty native plant seedlings were evenly planted in a 4 × 5 grid, with plants 20 cm from each other in each direction. For 10-species assemblages, two seedlings per species were planted and the positions of the seedlings were randomly located in the 4 × 5 grid of plants with the same spacing between neighboring individuals as that in monocultures.

To reduce nearby alternatives for pollinators and to increase the overall attractiveness of our common garden experiment, the flowers of all other flowering species within ≈ 20 m of the plots were continuously removed during the experiment. We watered the plots each evening during our sampling period. Only Achillea, Gaillardia, and Linum flowered during the experiment, and they each attracted substantial numbers of insect pollinators.

Data collection

We conducted five-minute observations of some combination of 43 different plots on 10 different days between June 02, 2015, and June 26, 2015, depending on what plants were flowering, for a total of 180 observations. Our total observation time (900 min) was similar to that reported in other studies (e.g., Muñoz and Cavieres, 2008; Lara-Romero et al., 2019; Albor et al., 2019), and sampling times per plot ranged from 10 to 50 total minutes with a mean of 22 min. The number of days sampled is included in our statistical model. For sampling, we chose the period of time during which flowering was common enough to find both one-species and two-species plots on the same day. Sampling was conducted only on days that were conducive to insect flight; with no sampling on days that were cooler than 12.8 °C or that were excessively windy (> 5.8 m/s). Plots were predominantly (80%) sampled on days that were sunny (8 of the 10 days) with temperatures ranging between 15.6 and 28.9 °C. The average sampling temperature was 25.6 °C. Our sampling was opportunistic as the species in flower in particular plots on a given sampling day varied substantially over the sampling period. Because of this natural variation, our sampling intensity per plot varied among the 43 plots, with the average number of sample days for a plot being 4.5 days, but ranging from 2 (a total of 10 min) to 10 (a total of 50 min) observation days per plot. No plot was sampled more than once per day. Observations were made between 0900 h and 1100. In each plot, we recorded the density of individual plants of each species in bloom. Over the 5-min observation period, all visits to each plant species in flower in the plot by all pollinators were recorded, which is by far the most common metric in such studies (Braun and Lortie, 2019). Landing on a flower was considered a visit. We recorded visits only when an insect landed on a flower in the plot. We also recorded the duration of visits when the density of pollinators was low enough to allow us to focus on a single insect. The exotic A. mellifera (honey bee) was identified to species, but otherwise pollinators were identified to either genus or family (Online Appendix, Fig. 1). This allowed us to measure the number of visits and some degree of the taxonomic richness of the pollinator community in each five-minute observation period. Taxonomic richness was defined as the number of different pollinator taxa that visited each plot in a given five-minute observational period.

We randomly collected 12 seed heads (mature inflorescences) from one randomly sampled Linum plant from each plot in which Linum occurred on June 29, 2015. These plots were classified as one-species and two-species plots. The seed heads were allowed to dry for 72 h, and then seeds were separated from the seed head. We counted the seeds in each of 12 seed heads from each Linum plant using a Konus dissecting scope (Model #5424).

Data analysis

We calculated the mean number of visits (calibrated by the sampling effort for each plot) of all pollinators to each species when flowering alone, and compared this to the mean number of visits to each species when flowering with one of the other two species. These means were tested with one way ANOVAs followed by Tukey post hoc tests. We measured the number of visits by all pollinator taxa to a plot during each of the five-minute observation periods (a plot was never sampled more than once per day) and tested the means of these between all one-species and two-species plots with an ANOVA with the number of species in flower (1 vs. 2) as a fixed variable, the density of all plants in flower as a covariate, and number of observation periods (days) that the plot was sampled as a covariate. We also recorded the number of different pollinator taxa that visited a plot during each of the five-minute observation periods (one per day) and tested the means of these between all one-species and two-species plots with the same ANOVA model. We recorded the mean duration of visits for as many pollinators as possible in one-species and two-species plots and divided these into two datasets: one for A. mellifera and Bombus species, and one for all other species, and tested these in separate ANOVAs with the same factors as above. This approach separated the social species from the solitary species. The rationale for this is that individuals of social species might forage differently than individuals of solitary species based on resource needs, individual learning, and the sharing of resource information possessed by social bees (Heinrich, 1976; Nieh et al., 2003; Dornhaus, 2006; Williams and Tepedino, 2003). The total pollinator taxa richness levels in one-species and two-species plots were also tested with the same ANOVA model. We compared the number of seeds per mature seed heads for Linum between one-species and two-species plots using an ANOVA with flowering species richness in plots as a fixed factor with seed head nested within plot as a random factor.

Results

Facilitation among plants

Each of the three plant species showed significant increases in average number of pollinator visits when grown with another species (Fig. 1). When grown in monoculture, Linum received an average of 1.6 ± 0.2 visits per sampling period, but when grown in plots with either Achillea or Gallardia, the visitation rate to Linum increased to 3.0 ± 0.8 and 3.9 ± 0.7 visits, respectively, per sampling period (Fspecies combination = 9.62; P = 0.001; Tukey for Linum alone vs. Linum with Achillea P = 0.021; Tukey for Linum alone vs. Linum with Gallardia P = 0.001). We could not compare the effect of Achillea on Gallardia because of low sample size, but Gallardia was visited an average of 2.6 ± 0.4 times when growing without another species in flower, versus 4.4 ± 0.5 times when growing with Linum in flower, an increase of 69% (Fspecies combination = 7.53; P = 0.015). Likewise, we could not measure the effects of Gallardia on Achillea because of low sample size, but Achillea received an average of 1.0 ± 0.6 pollinator visits during each observational period when growing alone, versus an average of 4.7 ± 0.3 pollinator visits when growing in plots with Linum in flower, an increase of 3.7 times the rate when growing alone (Fspecies combination = 14.83; P = 0.004). Our sample size for plots with Achillea and Linum together was very low, (n = 3), but each of these was sampled an average of five times over the flowering season.

Fig. 1
figure 1

Mean (± 1 SE) number of pollinator visitors to Linum lewisii, Gaillardia aristida, and Achillea millefolium when grown alone and with another flower species. Different letters indicate significant differences between means (see Table 1)

Full size image

For pollinators measured at the plot level, two-species plots affected pollinators differently than single-species plots, when all species combinations were combined. Two-species plots experienced a 66% increase in average visits over single-species plots (Fig. 2, Table 1). Neither the total density of plants in flower nor the number of observation days were correlated with the average number of visits per observation period (Table 1). Mean pollinator richness was 84% greater in two-species plots than in one-species plots, and neither the total density of plants in flower nor the number of observation days were correlated with average pollinator richness per observation period (Fig. 2, Table 1). Total pollinator richness across the entire sampling period was 67% higher in two-species plots than in one-species plots, and not surprisingly, total pollinator richness was affected by the number of observation days a plot received (Fig. 2, Table 1). Again total pollinator richness did not correspond with total flower density (Table 1).

Fig. 2
figure 2

Means (± 1 SE) for a number of visits, b average pollinator species richness, and c total pollinator species richness in plots with one species in flower versus plots with two species in flower. Different letters indicate significant differences (see Table 1)

Full size image
Table 1 Results for ANOVAs for the average number of visits for all pollinator taxa per each five-minute observation period of a given plot, the mean pollinator species richness attracted during each observation period, and the total pollinator species richness over all observations
Full size table

Seed set in Linum varied from 0 to 10 seeds per seed head with an average of 4.8 seeds per seed head across all plots combined. The average number of Linum seeds per seed head was 34% greater in two-species plots than in one-species plots (Fig. 3; Fspecies richness = 15.13; P < 0.0001).

Fig. 3
figure 3

Mean (± 1 SE) for number of seeds per Linum lewisii seed head in plots with Linum alone versus Linum with another flowering species. Different letters indicate significant differences at P < 0.05

Full size image

Pollinator community

We observed a total of eleven different pollinator taxa, that we could identify without destructive sampling, in our plots including the families Halictidae, Diptera, Vespidae, and Lepidoptera, the genera Megachile, Osmia, Hylaeus, Ceratina, Melissodes, Nomada, and Bombus, and the species Apis mellifera. Pollinators from Halictidae and Diptera, Megachile, and A. mellifera accounted for 86.1% of the visits (Online Appendix, Fig. 1).

As a group, A. mellifera and Bombus spp. (the two eusocial taxa, and the only two taxa that were exotic or with exotic species) showed no difference in visit duration between treatments (Fig. 4, Table 2). However, the group of all other pollinator taxa combined (all solitary natives) showed a 94% increase in visit duration in two-species plots (Fig. 4, Table 2). Visit duration showed no significant relationship with either flowering plant density or sampling intensity per plot (Table 2).

Fig. 4
figure 4

Mean (± 1 SE) pollinator visitation, or “visit duration” for plots with one flowering species versus plots with two flowering species. Pollinators are separated into a social insects, Apis mellifera and Bombus species, and b and nonsocial species. Different letters indicate significant differences (see Table 2)

Full size image
Table 2 Results for ANOVAs for the mean duration of visit for Apis mellifera and Bombus species averaged over all observation periods for each plot, and the mean duration of visit for all other pollinator species averaged over all observation periods for each plot
Full size table

Discussion

Our most salient finding was that, when visitation was measured to each species, three native intermountain prairie plant species experienced indirect facilitative interactions mediated by pollinators when any two of the three species were grown together. This facilitation was not mediated by an obvious magnet species because visitation per plant was not substantially different among any of the plant species. When flowering alone, Linum received 1.51 insect visitors per plant, Gallardia received 1.47 insect visitors per plant, and Achillea received 1.35 insect visitors per plant. Furthermore, the least visited species when alone, Achillea, significantly increased visitation to Gaillardia. All plant species experienced increased pollinator visitation when grown with another species regardless of the identity of the other species. Correspondingly, Linum plants produced more seeds per flower head when growing with either Achillea or Gaillardia than when alone in plots. Thus, our results support the idea that the diversity of flowering species in local patches, at a very fine grain (Braun and Lortie, 2019), can be important to plant–pollinator interactions in ways that others have found for magnet species (Molina‐Montenegro et al., 2008; Laverty, 1992; Johnson et al., 2000). This played out when visitation was measured at the whole-plot level, as plots with two species in flower received more visitations than plots with one species.

Not only did plant species increase visitation rates and seed set for each other, flowering species in pairs had positive effects on the pollinator community as well, when visitation was measured on a plot basis. Two-species plots increased the average pollinator diversity in a daily observation period, and the total pollinator diversity measured over the course of the experiment. Native pollinator species also increased the duration of visit in two-species plots. This may be important as wild native pollinators have been found to play a crucial role in agricultural seed set independently of the effects of managed pollinators (Garibaldi et al., 2013).

It is important to note that our small plots were only 30 cm apart, and although these distances are small, such a scale is a reasonable estimate of natural scales for flower patches. However, scale can have strong effects on the indirect effects of pollinators on plant interactions (see Albrecht et al., 2016), and depending on the mechanism, our small scale may increase the potential to show facilitation among plant species via pollinators. If complimentary attraction was important, a diverse patch very close to a monoculture patch might much more easily entice a monoculture’s pollinators than a diverse patch that is far away, simply because they are more easily detected. Thus, as distances increase among patches of different diversities, the facilitative relationship we found may be likely to wane (see review by Braun and Lortie (2019). Floral density, as opposed to our measurements of plant density, can also affect the balance of indirect facilitation mediated by pollinators (Underwood et al., 2020). Pollinator visitation rates should increase with floral density until flowers exceed the ability of pollinators to visit them—‘saturation’, which can drive shifts from indirect facilitation to competition among flowering plants (Rathcke, 1983). Importantly, the ambient density of pollinators can determine when facilitation shifts to competition with floral density (Ye et al., 2013).

There are at least four potential mechanisms for the indirect facilitation by pollinators we documented (see Braun and Lortie, 2019). First, it has been argued that convergent floral traits might promote facilitation via increased pollination among plant species (Rathcke, 1983; Schemske, 1981). This assumes that flower morphology has converged to the point that pollinators no longer distinguish between the species. This mechanism is unlikely to be important in our system because the three different plant species had very different floral characteristics.

Competitor-free space is a second mechanism for facilitation that occurs when pollinators move from a preferred floral resource to a less preferred resource upon the arrival of a superior competitor (Ghazoul, 2006). When this happens, the less preferred flowering species receives more pollinator visits. Many researchers have observed displacement of pollinators from flowers by other pollinators (Goulson, 2003; Johnson and Hubbell, 1974; Nagamitsu and Inoue, 1997), but this scenario is poorly studied and we observed very little displacement of initial pollinators by late comers.

Third, complimentary attraction promotes facilitation when each flowering plant species attracts a different group of pollinators. Thus, two or more species together will attract more visits than either of the species alone (Moeller, 2004; Rathcke, 1988). Flowering species in our plots attracted specific pollinators with differing degrees of success. For example, Gaillardia was very attractive to A. mellifera, attracting 56% of all honey bees observed. In contrast, Achillea accounted for only 7% of A. mellifera visits. Sweat bees and flies visited Achillea with regularity (24% and 25% of the observed sweat bees and flies, respectively). Thus, a plot that contained Gaillardia and Achillea would be likely to attract a richer total group of pollinators than single species. This may, in turn result in greater pollination success for both species if the pollinators move more frequently to nearby conspecifics. This is a likely mechanism for our patterns.

The fourth possible mechanistic explanation for indirect facilitation such as we observed, and a second more likely mechanism for our patterns, is resource complementarity. Resource complementarity occurs when different species within a plant community provide unique sets of resources. Thus, a pollinator that visits different species obtains a wider set of resources required for survival and reproduction. This differs from complimentary attraction in that the pollinator visitor is returning to flowers with known nutritional value rather than selecting flowers based on plant attractants such as color, shape, or scent. Pollinator choice due to resource complementarity may occur because different plant species produce different amino acids and carbohydrate mixtures in their pollen and nectar (Petanidou et al., 2006; Szczêsna, 2006). Insect pollinators that feed on a mixed diet have increased longevity and brood productivity than those deprived of diversity in their diet. Honey bees, for example, require 10 amino acids to complete brood development (De Groot, 1953). Herbert et al. (1970) showed that nurse bees fed only dandelion pollen (Taraxacum officinale) failed to rear brood; however, once dandelion pollen was supplemented with the missing amino acids, full brood development occurred. Alaux et al. (2010) found that A. mellifera colonies that fed on a mixed diet showed an increase in immunocompetence compared to colonies that fed on monocultures. Mevi-Schütz and Erhardthad (2005) found similar results for butterflies. Thus, for many insects, resources from many plant species may be necessary to meet life history requirements. There has been no studies of the amino acids content of the nectar or pollen for the species that we used. However, Petanidou et al. (2006) reported evidence that species in the Asteraceae family do not provide all of the amino acids needed for honey bee brood development. Crane (1975) and Galetto and Bernadello (2003) found that species in the Asteraceae and Linaceae produce very low sugar ratios in their nectar. Thus, the individual species that we tested may not provide either enough sugar or amino acids for the optimal performance of pollinators, and thus, mixed diets may be crucial.

In summary, we found evidence for indirect facilitation among three species and no evidence for competition for pollinators. Such facilitation may provide more complete suite of plant-supplied resources to pollinators. Furthermore, our results are consistent with a number of studies (see review by van der Plas, 2019), indicating that plant species diversity increases the overall species richness and diversity of the insect pollinator community.