INTRODUCTION North American native pollinators have been declining in recent decades (Buchmann and Nabhan 1996, Potts et al. 2010, Cameron et al. 2011, Burkle et al. 2013, Goulson et al. 2015). This decline potentially impacts not only the persistence of native plant species (Fontaine et al. 2005), but also the crucial agricultural services these insects provide; fruit and vegetable pollination by native insects in the United States alone has been valued at just over 3 billion dollars (Losey and Vaughan 2006). This service depends on the persistence of high-diversity pollinator assemblages, which can better withstand short-term environmental fluctuations as well as long-term stressors such as climate change (Kremen et al. 2007, Bartomeus et al. 2013).

Threats to pollinator diversity include the increased use of pesticides, the spread of exotic species, climate change, and, in particular, habitat loss and fragmentation (Rathcke and Jules 1993, Memmott et al. 2007, Brown and Paxton 2009, Whitehorn et al. 2012, Arbetman et al. 2013, Goulson et al. 2015). Habitat loss and fragmentation are the processes by which habitat patches become smaller and more isolated from each other (Wilcove et al. 1986). These changes in landscape structure can reduce species persistence and community diversity in pollinators as well as in their plant mutualists (MacArthur and Wilson 1967, Rathcke and Jules 1993, Brown and Paxton 2009, Öckinger et al. 2010). Although there is substantial evidence indicating that higher-diversity plant communities support greater pollinator diversity and vice versa (Potts et al. 2003, Fontaine et al. 2005, Ebeling et al. 2008, Blüthgen and Klein 2011), the mechanisms by which changes in landscape structure affect pollinator diversity remain poorly understood. Specifically, it is unclear whether habitat loss reduces pollinator diversity directly, or whether pollinator diversity is more strongly influenced by landscape-induced reductions in plant diversity. Additional insight into this topic might help to better inform management decisions regarding native pollinator conservation.

Here, we investigate the importance of the direct effects of landscape structure on pollinator diversity relative to indirect effects mediated by landscape-induced changes in plant species richness. We examine (1) the relationship between landscape spatial structure and pollinator diversity in insular grasslands, and (2) the influence of plant species richness on this relationship. Plants and pollinators differ in the distances they are able to disperse (Howe and Smallwood 1982, Cheplick 1998, Osborne et al. 1999, Nathan and Muller-Landau 2000, Gathmann and Tscharntke 2002, Knight et al. 2005, Pasquet et al. 2008), suggesting that the two groups might respond differently to landscape structure (Burkle and Alarcón 2011). We hypothesized that plants, being less mobile than pollinators, would be more sensitive to landscape structure and would affect pollinators via their mutualistic relationship.

METHODS

RESULTS The observed number of pollinator morphoguilds ranged from 0 to 13 per site, with a median of five morphoguilds. In total, we observed 28 butterfly morphoguilds; six of which were categorized, as large (>100 mm), nine as medium (∼75–100 mm), 10 as small (∼50–75 mm), and three as extra-small (<50 mm). Common butterflies included members of the genera Vanessa [Fabr.], Phyciodes Hübner, and Colias [Fabr.], and members of the family Hesperiidae.

The remaining 36 morphoguilds included solitary bees, hoverflies, bumblebees, honeybees, and wasps. We categorized 14 of these morphoguilds as large (>20 mm), nine as medium (∼10–20 mm), seven as small (∼5–10 mm), and six as extra-small body sizes (<5 mm). The bee community included representatives of genera such as Lasioglossum Curtis, Megachile Latreille, Xylocopa Latreille, and Bombus Latreille. Apis mellifera L. was a dominant species in many glades. Hemipenthes sinuosa (Wiedem.) and Chrysanthrax cypris Meigen were two of the dominant hover fly species, and wasps in the genus Polistes Latreille were common.

Plant species richness ranged from 15 to 53, with a median species richness of 35. The species richness of biotically pollinated plants ranged from 14 to 44, with a median species richness of 30. Warm season grasses such as Panicum virgatum L., Schizachyrium scoparium Nash, and Sorghastrum nutans Nash dominated the plant community. The glade flora also included a diverse assortment of forbs, with a large number of species in the Asteraceae, Fabaceae, and Lamiaceae families.

The relationships between landscape structure and plant richness (r = 0.36, p = 0.04, df = 30), and between plant richness and pollinator morphoguild richness (r = 0.38, p = 0.03, df = 31) were both statistically significant (p ≤ 0.05). Pollinator morphoguild richness was not significantly correlated with landscape structure (r = 0.20, p = 0.28, df = 30) or floral abundance (r = 0.03, p = 0.85, df = 30). The relationship between the richness of biotically pollinated plants and pollinator morphoguild richness (r = 0.39, p = 0.03, df = 31) was similar to that of total plant species richness. The relationship between landscape structure and biotically pollinated plant richness was not significant, but showed a positive trend (r = 0.32, p = 0.07, df = 30).

Our final SEM (Figure 1) contained significant paths between landscape spatial structure and plant species richness and between plant species richness and pollinator morphoguild richness. The path from proximity index to pollinator richness was not significant and was removed from the final model. Our final model contained an indirect relationship between landscape spatial structure and pollinator morphoguild richness, with plant species richness functioning as an intermediary (Figure 1). The structure of this model did not change when wind-pollinated species were excluded from the plant species richness values (Figure 2), although the relationship between landscape structure and plant species richness became marginally significant (p = 0.05 when excluding wind-pollinated species vs. p = 0.03 when all plant species were included). Survey date, time, cloud cover, and wind speed were not significantly correlated with plant richness, pollinator richness, or landscape structure, and were not significant when incorporated into the SEM.

Figure 1. 

Structural equation model of landscape spatial structure (habitat area and connectivity), plant species richness, and pollinator morphoguild richness. Arrows are labeled with standardized path coefficients and represent directional links between variables. Solid arrows denote significant relationships (p ≤ 0.05) and dashed arrows denote nonsignificant relationships (p > 0.05).

Figure 2. 

Structural equation model of landscape spatial structure (habitat area and connectivity), the species richness of biotically pollinated plants (excludes wind pollinated graminoid species), and pollinator morphoguild richness. Arrows are labeled with standardized path coefficients and represent directional links between variables. Solid arrows denote significant relationships (p ≤ 0.05) and dashed arrows denote nonsignificant relationships (p > 0.05).

DISCUSSION We found positive relationships between landscape structure and plant species richness, and between plant species richness and pollinator morphoguild richness. Landscape structure, however, was not directly correlated with pollinator richness, suggesting that in Ozark glades, plant communities are more sensitive to landscape structure than are pollinator communities. Our results suggest that although landscape structure does not directly influence pollinator diversity, landscape effects on plant species richness can lead to corresponding changes in the pollinator community. This indirect relationship between landscape structure and pollinator diversity highlights the importance of accounting for indirect effects in studies of diversity across trophic levels.

In Ozark glades, higher-diversity plant communities support higher-diversity pollinator communities and vice versa (Potts et al. 2003, Fontaine et al. 2005, Ebeling et al. 2008, Blüthgen and Klein 2011). However, because we were interested in the effects of landscape structure on this mutualism, we designed our structural equation model with a directional path from plant species richness to pollinator morphoguild richness and not the other direction. We reasoned that because landscape structure was not directly correlated with pollinator morphoguild richness, any influence of landscape structure on the plant-pollinator mutualism had to result from landscape-induced changes in plant species richness.

Pollinator diversity was not correlated with floral abundance. This result supports prior research indicating that floral diversity might be more important than floral abundance in supporting diverse pollinator communities (Steffan-Dewenter and Tscharntke 2001, Potts et al. 2003, Potts et al. 2006). Greater floral diversity might increase the number of available foraging niches via a greater diversity of floral characteristics, thereby facilitating visitation by a higher diversity of pollinators. Potts (2003) found evidence for this pattern, noting that the variety of nectar-foraging resources increases with plant diversity, and that nectar diversity was the best predictor of bee diversity. Junker et al. (2013) reported specialized foraging behavior in response to unique floral traits in every pollinator species observed, demonstrating pollinators' affiliations with particular floral niches. These findings are consistent with the idea that habitats with higher floral diversity support higher numbers of specialist pollinator species with unique foraging niches, as documented by Burkle and Knight (2012), who found greater plant and pollinator richness, as well as a higher proportion of specialist pollinators, in larger Ozark glades.

The SEM that excluded wind-pollinated plant species had the same overall structure as the original model (Figure 2), however, the r² values and standardized path coefficients were slightly lower in the former (Figure 2). This result is unexpected, because excluding nonnectar-providing plant species should help elucidate a functional relationship between pollinators and their resources. Our results suggest that greater plant richness might support higher diversity pollinator communities through an additional mechanism besides nectar-resource provisioning. For example, wind-pollinated species might contribute to pollinator habitat quality by providing non-foraging resources such as nesting sites or materials. Nesting sites and materials are known to be important determinants in pollinator community composition (Potts et al. 2005), although their importance relative to nectar resources has rarely been quantified. There is evidence to suggest that graminoid species in particular can be an important source of nesting resources. Graminoid species often have hollow or pithy stems, ideal nesting habitat for pollinators such as mason bees from the family Megachilidae, which prefer to nest in preexisting structures (Potts et al. 2005). Svensson et al. (2000) also found withered grasses and tussocks to be the favored nesting habitat in a study of eight species of bumblebees.

Plants appeared to be more sensitive than pollinators to landscape spatial structure. One explanation could be that the woodland matrix surrounding glades presents a more hostile environment to plant dispersal than to pollinator dispersal. Grassland plants generally cannot grow in the dense shade of woodlands, and wind-dispersed grassland seeds have been shown to travel poorly through forested areas (Damschen et al. 2014). According to Westrich's (1996) theory of partial habitats, pollinators use multiple habitat types due to diverse resource requirements (such as food, nesting sites, and nesting materials) that cannot be supplied by a single habitat alone. It might be, therefore, that pollinators' increased mobility relative to plants in the woodland matrix is not only due to their increased ability to traverse the landscape (Burkle and Alarcón 2011), but also to their ability to use resources found in this partial habitat. This idea is supported by research documenting pollinators along a wide gradient of canopy cover, with many species performing vital life history functions in the shelter provided by trees (Tscharntke et al. 1998, Grundel et al. 2010).

Although our results lend additional support to the importance of maintaining multiple partial habitats to meet pollinators' diverse resource requirements (Westrich 1996), the relative quantities of each partial habitat are also worth consideration. Our finding that the amount of woodland matrix present in the glade landscape does not directly affect pollinators suggests that the quantity of this partial habitat might have exceeded the threshold required by the pollinator community. Increased plant richness associated with well-connected landscapes appeared to have a positive influence on pollinator diversity; this suggests that resources available in glades might be more limiting to pollinator diversity than resources (such as nesting sites or materials) found in the woodland matrix. Our results, therefore, underscore the importance of maintaining large, well-connected glade habitats to provide pollinator communities with these limiting resources.

Our results indicate that diverse pollinator communities require diverse plant communities and that diverse plant communities require large well-connected habitat areas. However, urban and agricultural landscapes pervade much of North America at present, and the pollinator habitat remaining within these landscapes is highly fragmented. Our observation that pollinator morphoguild richness appears to be primarily limited by plant species richness, not landscape structure, indicates a potential for pollinator conservation within fragmented landscapes, provided that habitat patches are managed to maintain sufficient levels of plant diversity. Kearns and Oliveras (2009) provide one example of how the maintenance of natural habitat in and around an urban landscape can lead to stable pollinator communities over 100 yr, despite drastic land use-changes. However, the feasibility of this strategy at the scale of an agricultural landscape, where much pollinator habitat is currently being lost, remains uncertain. Therefore, our results primarily emphasize the importance of conserving and restoring the large contiguous habitat patches needed to support diverse plant communities, which in turn sustain diverse pollinator communities.