Introduction
Bees are important contributors to pollination services, but are currently facing a range of threats. Many bee species currently face population declines stemming from several different processes, including a low, discontinuous supply of floral resources, disease, habitat fragmentation, and climate change (Potts et al., Reference Potts, Biesmeijer, Kremen, Neumann, Schweiger and Kunin2010; Cameron et al., Reference Cameron, Lozier, Strange, Koch, Cordes, Solter, Griswold and Gene Robinson2011; Giannini et al., Reference Giannini, Acosta, Garófalo, Saraiva, Alves-Dos-Santos and Imperatriz-Fonseca2012; Hung et al., Reference Hung, Ascher, Gibbs, Irwin and Bolger2015; Scheper et al., Reference Scheper, Bommarco, Holzschuh, Potts, Riedinger, Roberts, Rundlof, Smith, Steffan-Dewenter, Wickens, Wickens and Kleijn2015). Bees and bee diversity benefit the pollination of crop and non-crop plants, thus it is critical to understand the factors that drive bee abundance and richness (Klein et al., Reference Klein, Steffan-Dewenter and Tscharntke2003; Breeze et al., Reference Breeze, Bailey, Balcombe and Potts2011; Winfree et al., Reference Winfree, Gross and Kremen2011). Habitat loss and change across landscapes can cause changes in plant reproductive success, although in some habitats or landscapes some of these effects may be mitigated through landscape management techniques (Harrison & Winfree, Reference Harrison and Winfree2015).
Urban gardens can provide semi-natural habitat that may act as a refuge for biodiversity, including bees (Goddard et al., Reference Goddard, Dougill and Benton2010; Tanner et al., Reference Tanner, Adler, Grimm, Groffman, Levin, Munshi-South, Pataki, Pavao-Zuckerman and Wilson2014). The area in urban gardens often determines the amount of green space in many urbanized cities, and in some cities, urban gardens cover between 23 and 36% of the city area (Gaston et al., Reference Gaston, Warren, Thompson and Smith2005; Loram et al., Reference Loram, Tratalos, Warren and Gaston2007; Mathieu et al., Reference Mathieu, Freeman and Aryal2007; Cameron et al., Reference Cameron, Blanuša, Taylor, Salisbury, Halstead, Henricot and Thompson2012). Urban gardens support many local, landscape, and socio-political features that may conserve biodiversity. For instance, local features such as mulch cover and flowering plant species richness augment spider activity and richness (Otoshi et al., Reference Otoshi, Bichier and Philpott2015). Garden size and socio-economic status of gardeners are crucial components for promoting avian richness and plant diversity (van Heezik et al., Reference van Heezik, Freeman, Porter and Dickinson2013). Further, urban gardens provide floral and nesting resources that may benefit insects (Wojcik et al., Reference Wojcik, Frankie, Thorp and Hernandez2008). Individual gardens may strongly differ in management techniques and thus in vegetation and insect composition (Loram et al., Reference Loram, Tratalos, Warren and Gaston2007). For bees in particular, carefully planned garden designs, including floral abundance, plant species richness, and appropriate plot sizes can support bee diversity and bee habitat (Frankie et al., Reference Frankie, Thorp, Schindler, Hernandez, Ertter and Rizzardi2005; Samnegard et al., Reference Samnegard, Persson and Smith2011; Baldock et al., Reference Baldock, Goddard, Hicks, Kunin, Mitschunas, Osgathorpe, Potts, Robertson, Scott, Stone, Vaughan and Memmott2015). Urban gardens are a key component of bee conservation because they can be managed for continuous floral resources (Threlfall et al., Reference Threlfall, Walker, Williams, Hahs, Mata, Stork and Livesley2015). Currently, however, there is a dearth of information about how the specific features of garden design influence bee communities (Wojcik et al., Reference Wojcik, Frankie, Thorp and Hernandez2008). There is also a lack of specific information about how the abundance of one common introduced species (Apis mellifera) is influenced by garden design, despite its ubiquity in human dominated landscapes, including urban landscapes in much of the world (e.g. Tommasi et al., Reference Tommasi, Miro, Higo and Winston2004, Matteson et al., Reference Matteson, Ascher and Langellotto2008, Frankie et al., Reference Frankie, Thorp, Hernandez, Rizzardi, Ertter, Pawelek, Witt, Schindler, Coville and Wojcik2009).
Understanding the diversity and distribution of flowers, an important bee resource, may contribute to understanding bee communities and conservation in urban landscapes. In general, understanding spatial connectivity can help predict species distribution, species persistence and migration (Moilanen & Nieminen, Reference Moilanen and Nieminen2002). Further, the spatial distribution of resources (e.g. clustering, size, patchiness) influence animal foraging behaviour, species richness, and species composition (Goulson, Reference Goulson1999; Ribas et al., Reference Ribas, Sobrinho, Schoereder, Sperber, Lopes-Andrade and Soares2005; Braaker et al., Reference Braaker, Ghazoul, Obrist and Moretti2014). For bees specifically, diversity, abundance, composition, and spatial distribution of floral resources affect bee foraging behaviour, abundance, species richness, and community composition and thus may strongly affect interactions between pollinators and plants (Torné-Noguera et al., Reference Torné-Noguera, Rodrigo, Arnan, Osorio, Barril-Graells, Correia Da Rocha-Filho, Bosch and Ollerton2014; Harrison & Winfree, Reference Harrison and Winfree2015). At very local scales, bee visitation rates to flowers can differ with floral resource patch size (Sih & Baltus, Reference Sih and Baltus1987) or with the presence of other plant species in the same habitat patch (Thomson, Reference Thomson1981). At larger spatial scales, visitation rates to flowers may be influenced by floral connectivity in a landscape (Torné-Noguera et al., Reference Torné-Noguera, Rodrigo, Arnan, Osorio, Barril-Graells, Correia Da Rocha-Filho, Bosch and Ollerton2014). Patchy (Hines & Hendrix, Reference Hines and Hendrix2005), and heterogeneous spatial resources across a landscape may allow foraging bees to switch to different floral resources and increase offspring production (Williams & Kremen, Reference Williams and Kremen2007). Yet, in some circumstances, floral diversity, rather than floral density drives bee foraging and as such, understanding the specific factors that drive bee population and diversity are important to increase pollination services (Jha & Kremen, Reference Jha and Kremen2012).
In this study, we examined floral resources and bee communities in urban gardens to determine how floral abundance, floral diversity, and floral spatial distributions within urban gardens are associated with changes in bee richness and abundance. Specifically, we tested the responses of the bee community to changes in floral resources with four response variables: abundance of all bees (hereafter bee abundance), abundance of Apis mellifera (hereafter A. mellifera abundance), species richness of all bee species (hereafter bee species richness), and diversity of all bee species (hereafter bee diversity). We investigated two main research questions: (1) Does floral abundance and diversity in gardens correlate with bee abundance, A. mellifera abundance, bee species richness, and bee diversity? (2) Does the spatial distribution or connectivity of floral resources within gardens influence bee abundance, A. mellifera abundance, bee species richness, and bee diversity? We hypothesized that increases in floral abundance and diversity and more clustered floral resources would result in increases in bee abundance, A. mellifera abundance, bee species richness, and bee diversity in urban gardens. We also examined the role of floral abundance and spatial distribution in relation to other local and landscape characteristics of urban gardens important for urban bee communities.
Methods
Study sites
Between July and early August 2015 we surveyed 18 urban gardens, ranging in size from 444 to 15,525 m2, across three counties (Monterey, Santa Clara, and Santa Cruz) in the California central coast (fig. 1). All gardens included vegetable patches that had been in regular cultivation for at least 5 years, and many also included various ornamental, native, and non-native plants. In the centre of each garden, we established a 20 × 20 m2 plot within which all sampling was performed.
Fig. 1. A map of the Central coast region of California showing the 18 urban garden sites in Monterey, Santa Clara, and Santa Cruz Counties, and land cover types in the study region and surrounding the garden study sites with three zoomed in panels to show (a) a garden surrounded primarily by urban and natural land, (b) a garden surrounded by natural, open, and urban land, and (c) a garden surrounded by primarily urban and agricultural land.
Bee surveys
We sampled bees with elevated pan traps and hand netting (Grundel et al., Reference Grundel, Frohnapple, Jean and Pavlovic2011). We constructed pan traps using 400 ml plastic bowls (yellow, white, and blue) painted with Clear Neon Brand and Clear UV spray paint. We placed pan traps over 3 days in early July, from approximately 8 AM until 7 PM on each day, and trapped bees were collected daily. We placed three 1 m tall polyvinyl chloride (PVC) pipes in the ground in a triangle formation, 5 m apart within each of the 20 × 20 m2 plots, and placed one bowl of each colour on top of PVC tubes (Tuell & Isaacs, Reference Tuell and Isaacs2009). We filled bowls with 300 ml of water and 4 ml of unscented Dawn dish soap. In addition, we sampled bees using aerial nets at each site, over the days of 7–9 July, 31 July, and 2 August 2015. We searched for and captured bees in nets for a total of 30 min per site. We netted bees that were observed on flowers, within 20 m of and inside the 20 × 20 m2 plots in each site. We stored all captured bees for later identification. We performed bee identifications with reference to online resources, image databases, books, and dichotomous keys (Roberts, Reference Roberts1973a , Reference Roberts b ; Michener, Reference Michener2007; Gibbs, Reference Gibbs2010; Frankie et al., Reference Frankie, Thorp, Coville and Ertter2014; Ascher & Pickering, Reference Ascher and Pickering2015; Packer, Reference Packer2015). We identified all specimens to the highest taxonomic level possible, and for more difficult specimens we allocated them to morphosopecies. We also compared our specimens to specimens held in the Kenneth S. Norris Center for Natural History on the University of California, Santa Cruz campus. All voucher specimens are housed in the Philpott Laboratory at the University of California, Santa Cruz.
Floral surveys
For floral surveys, we divided the 20 × 20 m2 plot into 100 2 × 2 m2 quadrats and assigned each quadrat a spatial coordinate (A-J, 1–10) for use in spatial analysis. Before counting flowers, we spent 30–45 min observing bees and noting all floral species being visited by bees in that site on that day. Then, in each quadrat, we counted or estimated floral abundance of species being visited by bees. Most flowers were exhaustively counted. For flower species where we estimated abundance, we counted the number of flowers on each of three inflorescences, took the average value, and then multiplied by the total number of inflorescences in the quadrat. We noted the colours of each flower (white, yellow, purple, red, orange, purple, or blue) and identified all flowering plants to species or morphospecies.
Site characteristics
To determine if local- and landscape-scale characteristics had an effect on bee species richness and abundance, we measured ground cover within our plots, and classified nearby land cover types surrounding each site. At the local scale, we noted the percent ground covered with bare soil, herbaceous plants, and mulch within four 1 × 1 m2 plots in our 20 × 20 m2 plot. At the landscape scale, we classified the land cover types within 2 km buffers surrounding each garden with data from the 2011 National Land Cover Database (NLCD, 30 m resolution) (Homer et al., Reference Homer, Dewitz, Yang, Jin, Danielson, Xian, Coulston, Herold, Wickham and Megown2015). We chose 2 km buffers as 1.5–2 km is the median maximum foraging range of bee species for which data exist (Zurbuchen et al., Reference Zurbuchen, Landert, Klaiber, Müller, Hein and Dorn2010). We created four surrounding landscape categories: natural habitat, open, urban, and agriculture by combining NLCD land cover classes. Our natural habitat area included deciduous (NLCD number 41), evergreen (42), and mixed forests (43), dwarf scrub (51), shrub/scrub (52), and grassland/herbaceous (71) and is the only landscape category with predominantly natural vegetation. Three of these categories (urban, open and agriculture) represent areas heavily impacted by humans, although they differ in the predominant ground cover. According to the NLCD descriptions (see Homer et al., Reference Homer, Dewitz, Yang, Jin, Danielson, Xian, Coulston, Herold, Wickham and Megown2015), urban areas (combining low [22], medium [23], and high-intensity developed land [24]) contain between 20 and 100% impervious surface; open areas (21) are vegetated mostly in the form of lawn grass; and agricultural areas (combining pasture/hay [81] and cultivated crops [82]) have at least 20% crop or pasture grass cover. We chose these four landscape categories based on knowledge of bee foraging and nesting needs from the literature. Other land cover types covered <5% of the total area and were not included. We assessed land cover with spatial statistics tools in ArcGIS v. 10.1.
Data analysis
To answer our two questions, we used four different response variables: bee abundance, A. mellifera abundance, bee species richness, and bee diversity. Bee diversity was calculated with the Shannon–Wiener index (H´). We pooled all pan trap and hand-netting data from each site for all analysis. We included floral abundance characteristics, floral distribution characteristics, other local factors, and landscape factors as explanatory variables in a single statistical model (see below). Floral characteristics included total number of flowers and flower species per site, mean number of flowers per quadrat, max number of flowers per quadrat, mean number of white flowers per quadrat, as well as the spatial distribution of flowers. Aside from floral resource distribution, all local and landscape factors included are known to affect bee species richness and abundance in our study sites (Quistberg et al., Reference Quistberg, Bichier and Philpott2016). To our knowledge, no study to date has looked at floral resource distribution as an additional predictor of bee communities in urban gardens. We found a large range in all measured variables in the different study sites (Supplementary Table 1). To calculate the spatial distribution of flowers, we mapped the 100 quadrats for each site and joined the floral resource data to each quadrat in ArcGIS 10.1. Then for each site, we used spatial statistics tools to calculate six nearest-neighbour ratios (NNRs) for each site based on data for quadrats with ≥15, ≥50, and ≥100 flowers, ≥15 white flowers, and ≥2 species of flowers. We chose floral abundance thresholds of 15, 50, and 100 flowers per quadrat because those corresponded to roughly 40, 20, and 10% of all quadrats sampled. We included quadrats with white flowers given their importance for urban bees in our sites (Quistberg et al., Reference Quistberg, Bichier and Philpott2016). NNR calculates spatial patterns, such as clustering and dispersion. A smaller NNR value indicates a higher degree of clustering. Thus our analysis included five floral abundance variables (total floral abundance in a site, total floral species richness in a site, the mean number of flowers per quadrat, mean number of white flowers per quadrat, and the max number of flowers per quadrat), five floral distribution variables (site-level NNR values for quadrats with ≥15, ≥50, or ≥100 flowers, ≥15 white flowers per quadrat, and ≥2 species of flowers per quadrat), three other local factors (percent ground cover with bare ground, herbaceous vegetation, and mulch), and four landscape variables (percent of landscape with open, natural, agricultural, or urban land use within 2 km) for 18 explanatory variables.
To check for correlation among explanatory variables, we ran Pearson's correlations. We divided explanatory variables into four groups: (1) floral abundance and richness, (2) floral spatial distribution, (3) other local factors, and (4) landscape factors, examined which variables were highly correlated (P < 0.01), and selected one of the correlated variables as a representative for subsequent analysis (see Supplementary Methods). The nine explanatory variables chosen for subsequent analyses were mean number of flowers per quadrat, total flower species richness, NNR for quadrats with ≥15 flowers, NNR for quadrats with ≥50 flowers, NNR for quadrats with ≥100 flowers, mulch cover, herbaceous cover, urban land cover in 2 km, and agriculture in 2 km.
We used generalized linear models (GLMs) with the glm function in R (R Development Core Team, 2014) to examine relationships between selected floral abundance and distribution variables, other local factors, landscape characteristics and bee abundance, A. mellifera abundance, bee species richness, and bee diversity. We tested all combinations of different variables with the ‘glmulti’ package (Calcagno & de Mazancourt, Reference Calcagno and de Mazancourt2010) and selected the top model based on the AICc values. For models where the AICc for top models was within 2 points of the next best model, we averaged models (up to the top ten models) with the MuMIn package (Barton, Reference Barton2012) and report conditional averages for significant model factors. As dependent variables were normally distributed, we used Gaussian error structure for GLMs (i.e. models were equivalent to multiple linear regression models), and report corrected Akaike Information criterion (AICc) values, P-values, and multiple linear model R 2 values for all best models. All residuals from the best models conformed to the conditions of normality as checked with QQ-Plots and Shapiro–Wilk tests.
Because of the potential for managed hives of A. mellifera to influence bee abundances, we compared bee abundance, A. mellifera abundance, bee species richness, and bee diversity in sites with and without known managed honeybee hives with t tests. Finally, we examined correlations between A. mellifera abundance and bee species richness and bee diversity with simple linear regressions.
Results
We collected 1354 bee individuals from 43 species. We collected 5 bee families; the most abundant family was Apidae representing 70% of total individuals captured. The most abundant bee species was A. mellifera (58% of individuals captured), followed by Halictus tripartitus (10.1%), Bombus caliginosus (4.4%), and Bombus vosnesenskii (1.5%).
Bee abundance, A. mellifera abundance, bee species richness, and bee diversity were most affected by urban land cover, floral abundance, and floral spatial distribution. The model that best explained bee abundance included only urban land cover within 2 km (table 1). Increasing urban land cover predicted lower bee abundance (P = 0.015, fig. 2a). The model that best explained A. mellifera abundance included urban land cover and NNR for quadrats with ≥15 flowers (table 1). A. mellifera abundance decreased with higher urban cover (P < 0.001, fig. 2b) and increased as floral resources became more patchy (P < 0.001, fig. 2c). The models that best explained bee species richness and bee diversity both included mean number of flowers in a quadrat and NNR for quadrats with ≥15 flowers (table 1). Bee species richness declined as floral abundance increased (P = 0.018, fig. 3a) and as floral resources became more patchy (P = 0.031, fig. 3b). Likewise, bee diversity declined as floral abundance increased (P = 0.014, fig. 3c), and as floral resources became more patchy (P = 0.003, fig. 3d). We also noted negative correlations between the abundance of A. mellifera and bee species richness (R 2 = −0.561, P < 0.05, fig. 4a) and bee diversity (R 2 = −0.715, P < 0.01, fig. 4b).
Fig. 2. Correlations showing relationships between percent urban land cover and (a) number of bee individuals and (b) number of Apis mellifera individuals and the nearest-neighbour ratio (NNR) for quadrats with ≥15 flowers and (c) number of A. mellifera for bees collected in urban gardens in the Central coast region of California. The lines show the best fit and the grey area cover confidence bands based on the generalized linear models. Smaller NNR values indicate stronger floral clustering.
Fig. 3. Correlations showing relationships between mean number of flowers per 2 × 2 m2 quadrat and (a) bee species richness and (c) bee diversity, and between the nearest-neighbour ratio (NNR) for quadrats with ≥15 flowers and (b) bee species richness and (d) bee diversity for bees collected in urban gardens in the Central coast region of California. The lines show the best fit and the grey area covers confidence bands based on the generalized linear models. Smaller NNR values indicate stronger floral clustering.
Fig. 4. Correlations showing relationships between the number of honeybees (Apis mellifera) and (a) bee species richness and (b) bee diversity for bees collected in urban gardens in the Central coast region of California. The lines show the best fit and the grey area covers confidence bands based on the generalized linear models.
Table 1. GLM results table showing all response variables, explanatory variables included in the best models, AICc values, residual degrees of freedom, and R 2 values for general linear models.
Discussion
We investigated the effect of floral abundance, distribution, and other local and landscape factors on bee communities and found that floral spatial distribution is one of the most important drivers of bee species richness, bee diversity, and A. mellifera abundance. In addition, floral abundance and urban land cover are important drivers of bee communities. Bee abundance was significantly negatively correlated with urban cover in the landscape, but not with other floral abundance or distribution factors, or ground cover characteristics. Habitat loss associated with urbanization is one main cause of bee declines (Martins et al., Reference Martins, Goncalves and Melo2013), and other studies have documented drops in bee abundance with increases in concrete, buildings, and other types of impervious cover at the landscape level (Bates et al., Reference Bates, Sadler, Fairbrass, Falk, Hale and Matthews2011; Threlfall et al., Reference Threlfall, Walker, Williams, Hahs, Mata, Stork and Livesley2015). In addition, impervious surfaces limit nesting opportunities for bees and can increase bee foraging distances (Fortel et al., Reference Fortel, Henry, Guilbaud, Guirao, Kuhlmann, Mouret, Rollin and Vaissière2014). In our study, natural and open land cover negatively correlated with urban land cover, thus these variables, which were excluded from the analysis, may also impact bee abundance positively. Therefore, declines in urban developed cover and increases in cover by natural habitats (e.g. forest and grassland) likely both promote bee abundance, especially in areas with little natural habitat remaining (Winfree et al., Reference Winfree, Aguilar, Vázquez, LeBuhn and Aizen2009). For example, natural habitat provided by green roofs or small patches of ornamental plants can provide suitable habitat for bees to forage and collect floral resources (Tonietto et al., Reference Tonietto, Fant, Ascher, Ellis and Larkin2011; Garbuzov et al., Reference Garbuzov, Madsen and Ratnieks2015).
We found that the abundance of A. mellifera, by far the most common bee species collected in our study, declined with increases in urban cover, increased with more dispersed floral resources, but did not respond to other local factors. Increasing amount of urban cover is implicated in declines of bee abundance, generally (e.g. Potts et al., Reference Potts, Biesmeijer, Kremen, Neumann, Schweiger and Kunin2010). However, A. mellifera usually thrives in urban green spaces such as public parks and residential neighbourhoods, more so than other wild bees. This is likely because A. mellifera is a floral generalist, because wild bees may lack appropriate nesting habitat in urban areas (Threlfall et al., Reference Threlfall, Walker, Williams, Hahs, Mata, Stork and Livesley2015), and because honeybees are most likely managed and nests are provided for them. Although many studies note A. mellifera as the most common bee found in urban garden studies (e.g. Tommasi et al., Reference Tommasi, Miro, Higo and Winston2004; Matteson et al., Reference Matteson, Ascher and Langellotto2008; Frankie et al., Reference Frankie, Thorp, Hernandez, Rizzardi, Ertter, Pawelek, Witt, Schindler, Coville and Wojcik2009), none actually examine whether landscape features correlate with A. mellifera abundance within urban habitats. In addition, few studies have described floral spatial distribution as an important predictor for honeybees. We found that A. mellifera abundance was higher in sites with more patchy (i.e. less clustered) floral resources and this finding may provide insight for managing A. mellifera abundance in urban gardens. A. mellifera is a generalist species and its medium size permits it to forage large distances (Greenleaf et al., Reference Greenleaf, Williams, Winfree and Kremen2007), thus it is unlikely that A. mellifera abundance would be negatively affected by dispersed floral resources (Beekman & Ratnieks, Reference Beekman and Ratnieks2000). In other landscapes, A. mellifera abundances were positively associated with large landscape scales in landscapes with fewer semi-natural habitats, thus showing adaptation to more fragmented habitats and patchy resources (Steffan-Dewenter et al., Reference Steffan-Dewenter, Munzenberg, Burger, Thies and Tscharntke2002). Eusocial insects, such as A. mellifera, that live in large colonies recruit foragers to search for patches with abundant resources. One study reported the colony health or ‘energy status’ of A. mellifera influenced the foraging distance, for instance, when the floral resources were high A. mellifera foraged small patches and short distances, and when resources were low they foraged longer distances and larger patches (Schneider & McNally, Reference Schneider and McNally1993). Therefore, A. mellifera may be better equipped than other bees to experience spatial changes in floral resources because they forage at variable distances when floral resources are also variable. Finally, many urban sites, including gardens, may actively promote A. mellifera by maintaining managed hives. Of our 18 sites, four had managed hives at the time of our study, but we do not know if homeowners in private property surrounding other sites may have had hives. A. mellifera abundance was significantly higher (t test, P = 0.006) in the four sites with known managed hives, but there were no differences in bee species richness, bee diversity, or (non-A. mellifera) bee abundance (t tests, P > 0.05) in sites with and without known managed hives.
We found that changes in bee species richness and bee diversity were largely driven by floral abundance (but not landscape factors). While floral abundance is often associated with higher bee richness in urban areas (e.g. Matteson & Langellotto, Reference Matteson and Langellotto2010; Wojcik and McBride, Reference Wojcik and McBride2012; Hülsmann et al., Reference Hülsmann, von Wehrden, Klein and Leonhardt2015), we found that bee species richness and diversity was lower in sites with more flowers and patchier flower resources. This may be due to sampling effects whereby more flowers available result in fewer bees captured in pan traps. Our analysis examined mean number of flowers per quadrat, but this was also correlated with total floral abundance, maximum floral abundance per quadrat, and also with floral abundance of white flowers, so any of these variables may drive the observed effects.
In contrast to patterns for A. mellifera abundance, we found that sites with more clustered floral resources supported higher bee richness and bee diversity. This is a novel finding as the first study to assess how floral distribution within urban ecosystems impacts bee communities and potentially bee conservation. Others have documented increases in abundance of individual bee groups (e.g. bumble bees) in areas with patchy floral resources (Wojcik & McBride, Reference Wojcik and McBride2012), but have not examined entire communities. Clustered floral resources may support an array of bees that forage both short and long distances, but may be particularly important for smaller bees that exhibit limited foraging ranges (Zurbuchen et al., Reference Zurbuchen, Landert, Klaiber, Müller, Hein and Dorn2010). Further, different bees (even within the same genus) may respond differently to floral patch size (Sowig, Reference Sowig1989). The frequency of pollinator visits may decrease as flower patch size increases because searching for unvisited flowers in small patches may allow bees to optimize their foraging strategy (Goulson, Reference Goulson2000). Similarly, floral density effects are strong at low densities because plants facilitate one another's pollinator attraction, while higher floral densities tend to have weak pollinator attraction because plants compete for pollinator attraction (Essenberg, Reference Essenberg2012). Bee conservation in intensified agricultural systems (with low floral resources) can be bolstered by adding clumped spatial elements such as hedgerows or buffer strips (Klein et al., Reference Klein, Vaissière, Cane, Steffan-Dewenter, Cunningham, Kremen and Tscharntke2007). These additions likely work to augment bee diversity because bees in human-managed systems respond to clustered floral resources. For example, in a different agricultural system (tropical coffee systems), bee diversity did not respond to floral resources clumping at the field scale, but bee diversity increased in sites with branch and shrub scale floral clustering, thus emphasizing the notion that responses of bee diversity to floral clustering are dependent both on floral abundance but also on spatial scale (Veddeler et al., Reference Veddeler, Klein and Tscharntke2006).
One of the striking patterns found is that A. mellifera abundance and bee species richness and bee diversity responded to floral spatial distribution in opposite ways – with bee species richness responding positively to clustering, and A. mellifera abundance responding negatively to floral clustering. This prompts the question of whether interactions between A. mellifera and other bee species may be driving observed patterns. We posit that due to extensive foraging ranges and generalist preferences, A. mellifera could be foraging in dispersed floral patches, allowing smaller bees or other bee species to occupy the clustered patches of flowers. A. mellifera presence may restrict access by other bees through interference competition, or by apparent competition if A. mellifera deplete nectar resources driving other bees to search elsewhere (e.g. Schweiger et al., Reference Schweiger, Biesmeijer, Bommarco, Hickler, Hulme, Klotz, Kühn, Moora, Nielsen, Ohlemüller, Petanidou, Potts, Pyšek, Stout, Sykes, Tscheulin, Vila, Walther, Westphal, Winter, Zobel and Settele2010). Yet, there may be minimal interference of floral resources by honeybees compared to native bees because different bee groups may not share floral resources (Pedro & Camargo, Reference Pedro and Camargo1991). The assumed widespread effects of A. mellifera on other bees are often based on observations, but not long-term population assessments (Paini, Reference Paini2004); thus, careful consideration is necessary. Some studies have taken an experimental approach to examine the influences of removal of one numerically dominant bee on foraging patterns of other species. For example, removal of a numerically dominant bee (Bombus sp.) from alpine meadows in Colorado influenced the floral visitation of other pollinator species (Brosi & Briggs, Reference Brosi and Briggs2013). One experimental study demonstrated that in small and isolated flower patches, increased honeybee density reduced visitation rates, niche breadth, and reproduction of the red mason bee (Hudewenz & Klein, Reference Hudewenz and Klein2015). Another potential mechanism driving negative relationships between honeybees and other bees may be the transmission of disease from A. mellifera to wild bees (Furst et al., Reference Furst, McMahon, Osborne, Paxton and Brown2014). Regardless, any interactions between A. mellifera and other bee species may have important implications for pollination services in urban gardens (Greenleaf & Kremen, Reference Greenleaf and Kremen2006). A. mellifera thrives in urban settings (Tommasi et al., Reference Tommasi, Miro, Higo and Winston2004), but their high floral visitations have led to a reduction in the fitness of native bees and the flowers other bees pollinate (Gross & Mackay, Reference Gross and Mackay1998). For some plant species, honeybees have poor pollination efficiency and may create discrepancies between higher bee visitation rates and lower seed sets in urban sites (Leong et al., Reference Leong, Kremen and Roderick2014). Certainly, further research and experimentation in understanding interactions between native bees and A. mellifera is warranted.
Urban gardens are important in bringing environmental awareness about ecosystem services to human communities and for sustaining biodiversity of ecological communities (Goddard et al., Reference Goddard, Dougill and Benton2010). Urban gardens connect fragmented areas impacted by urbanization and intensified agriculture by linking floral communities, bee communities, and stewardship by the gardeners. Increasing urbanization and habitat loss puts significant pressures on these isolated gardens to support great diversity, thus it is crucial to study how to diversify urban systems to promote biodiversity (Philpott et al., Reference Philpott, Cotton, Bichier, Friedrich, Moorhead, Uno and Valdez2014). Our main findings show that abundance and spatial distribution of floral resources and landscape factors are important for maintaining diverse and abundant bee communities and could contribute to management decisions within urban gardens. Our results suggest that bee diversity responded positively to spatial aggregations of floral resources, and that spatial arrangement of flowers is important in managing urban habitats for bees. Thus, gardeners might strive to plant several smaller clumped flower patches. At larger scales, promoting natural and open space within urban areas may also encourage overall bee abundance, richness, and conservation and pollination services within urban landscapes.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007485317000153
Acknowledgements
We thank P. Bichier, H. Cohen, M. Egerer, M. MacDonald, J. Burks, R. Schreiber, and S.-S. Thomas for contributing to field data collection and data entry. Thanks to M. Bello and B. Hall for assistance with spatial analysis and GIS. The work was partially supported by the UC MEXUS-CONACYT grants for collaborative projects to S. Philpott and H. Morales, UCSC Committee on Research Faculty Research Grant to S. Philpott, and UCSC Heller Endowment Funds to S. Philpott.
