Pollination services are mediated by bee functional diversity and landscape context

Agriculture, Ecosystems and Environment 200 (2015) 12–20

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee

Pollination services are mediated by bee functional diversity and landscape context Kyle T. Martins 1, *, Andrew Gonzalez 1, Martin J. Lechowicz 1 Department of Biology, McGill University, 1205 Docteur Penfield Avenue, Montreal, QC H3A 1B1, Canada

A R T I C L E I N F O

A B S T R A C T

Article history: Received 12 May 2014 Received in revised form 17 October 2014 Accepted 21 October 2014 Available online xxx

Wild bees, which exhibit multiple functional traits enabling pollination of apples (Malus domestica Borkh), can potentially compensate for recent declines in domesticated honey bees (Apis mellifera Linnaeus) that are conventionally employed to ensure apple fruit and seed set. Whether compensation is possible will depend on functional diversity in the wild bee community and on the distribution of habitat and resources within the landscape surrounding an orchard that affect wild bee abundance. We studied pollination services and bee functional diversity in 20 apple orchards in southern Quebec, Canada. We evaluated pollinator efficacy by studying: apple visitation rates, approach (front or side-working), body size, foraging type (pollen or nectar foraging), sociality, temporal and climatic activity patterns, and pollen carrying habit. Pollination services were measured as apple fruit set and seed set. A distance-based measure of functional diversity, calibrated with bee traits and weighted by species relative abundance in the wild bee community, was used to model pollination services. We correlated the landscape composition and configuration of surrounding natural (forest) and semi-natural (meadow) habitats with bee diversity and pollination services. The incidence of fruit set and seed set in orchards increased with bee functional diversity. Complementarity between managed versus unmanaged bees in traits associated with foraging and resource use drove this relationship. Seed set was also negatively correlated with both the mean distance from surrounding meadows and the total area of surrounding orchards. Bee functional diversity was positively associated with surrounding meadow and forest area. These two land classes complement each other in their seasonal provision of foraging resources for bees. Our models can be used to prescribe management and conservation objectives for bee habitat management that promote pollination services. We identify useful wild bee pollinators and discuss their needs in terms of landscape composition and configuration. ã 2014 Elsevier B.V. All rights reserved.

Keywords: Agroecosystems Functional complementarity Honey bees Malus x domestica Pollination limitation Wild bees

1. Introduction Recent declines in domesticated honey bees (Apis mellifera Linnaeus) may jeopardize the sustainability of agro-ecosystems (Aizen and Harder, 2009; Neumann and Carreck, 2010). Wild insects are an important but underappreciated group of crop pollinators, representing half of all pollinator visits to crop flowers in agricultural systems worldwide (Garibaldi et al., 2013). Wild bees are derived from the landscapes surrounding farms (Kennedy et al., 2013) and freely pollinate crop flowers, providing ecosystem services that could help offset honey bee declines. Our study

* Corresponding author. Tel.: +1 514 398 6400; fax: +1 514 398 5069. E-mail addresses: [email protected] (K.T. Martins), [email protected] (A. Gonzalez), [email protected] (M.J. Lechowicz). 1 Tel.: +1 514 398 6400; fax: +1 514 398 5069. http://dx.doi.org/10.1016/j.agee.2014.10.018 0167-8809/ã 2014 Elsevier B.V. All rights reserved.

focuses on both the effect of the wild bee community on pollination services and the role played by landscape composition and configuration surrounding farms on the availability of wild bees. In a diverse assemblage of bees species can be characterized by particular functional traits that facilitate pollination services to a greater or lesser degree (Blüthgen and Klein, 2011). For instance, bee species can vary in the number of flowers visited per time unit and in their efficacy at depositing or removing compatible pollen on the stigma per visit (Thomson and Goodell, 2001), as well as in their physiological restrictions that limit their activity to certain abiotic conditions (Rader et al., 2013). When contributions to ecosystem function differ among species, traits from several species taken together can enhance overall functional performance levels (Blüthgen and Klein, 2011). Provided that bees in the community do not exhibit functional redundancy, there is the potential for additional enhancement through complementarity effects (Blüthgen and Klein, 2011). Hence, functional diversity in

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

the bee community can be anticipated to enhance pollination services, an expectation that has been validated both experimentally (Fründ et al., 2013) and by observational studies (Hoehn et al., 2008; Klein et al., 2008). Research on bee functional diversity is especially important for tree crops such as apple (Malus x domestica Borkh). Apple cultivars are mostly self-incompatible and require insect mediated crosspollination to set fruits and seeds. Poor pollination can reduce seed set, which leads to asymmetrical pomes (Brault and Oliveira, 1995; Sheffield, 2014) that fail to meet market expectations. Orchardists typically rent domesticated honey bees to ensure apple pollination (Free, 1993). Yet honey bees are not the most effective apple pollinators in that they “sidework”, i.e., rob nectar from apple flowers without contacting and thus fecundating the stigma (Schneider et al., 2002). Many wild bee species in orchards (Gardner and Ascher, 2006; Watson et al., 2011) have traits better enabling apple pollination: carrying more compatible pollen (Kendall, 1973), transferring pollen at a higher rate (Thomson and Goodell, 2001), having a stronger preference for Malus flowers (Kendall and Solomon, 1973) and compensating for honey bees under adverse environmental conditions (Boyle-Makowski and Philogene, 1985). Wild bees have the potential to complement honey bees in pollinating apples, enhancing fruit yield and quality. The role of wild bee diversity in apple pollination, however, has yet to be quantified, and wild bees are not commonly incorporated into orchard management. The availability of nesting and foraging resources for wild bees in the landscape surrounding farms will influence the degree to which farmers can depend on wild bees to reduce the need to import honey bees. Wild bees track resource availability from forests to agricultural crops and meadows throughout the year (Mandelik et al., 2012), such that the availability of floral resources in the landscape around a farm can influence crop pollination. For example, the abundance and diversity of wild bees visiting apple blooms increased with nearby forest cover (Watson et al., 2011). We build on previous work by sampling the diversity of all Apis and non-Apis bees in apple orchards situated along a gradient of forest cover and open field habitat. In general, little is known about the

13

importance of the composition and configuration of complementary habitat and resource patches in the landscape (Mitchell et al., 2013). By quantifying the relationship between bee diversity, apple pollination, and landscape metrics, we provide clear objectives for land use managers regarding how best to ensure the security of wild bee pollination services (Kremen et al., 2007). We investigated the relationships among multiple functional traits, bee diversity, pollination services and the landscapes features upon which wild bees depend. We used seed and fruit set as a proxy for pollination service delivery, since both variables are the most direct results of pollination (Dennis et al., 2003). If measured soon after the apple bloom, these variables are not confounded by local management decisions (e.g., apple thinning) that otherwise influence end-of-season crop yield estimates. We addressed two specific questions: (1) Is bee functional diversity correlated with apple fruit set and seed set? and (2) Which landscape characteristics best predict bee diversity and pollination services in apple orchards? 2. Methods We studied 20 apple orchards in the Montérégie administrative region of southern Quebec, Canada (Fig. 1). Sites were all situated along hillsides with good cold air drainage to minimize the effect of frost damage to apple blossoms. To test the representativeness of our orchard landscapes, we used a two-sample t-test to compare the amount of forests, scrubland, and meadows within 500 m of our study sites with those from a random sample of 108 orchards throughout the region. The mean percent natural land cover surrounding our orchards (40  17%) was not significantly different (t = 1.83, P > 0.05) from orchards across the Montérégie (32  19%), and is comparable to other apple producing regions (e.g., Watson et al., 2011; Marini et al., 2012). To secure the spatial independence of our sampling points, we spaced sites at least 1766 m apart, beyond the foraging range of most bee species in our area (Greenleaf et al., 2007). All the orchardists practiced integrative pest management in consultation with a professional agronomist. Fourteen of the 20 orchards rented honey bees at recommended

Fig. 1. Map of study orchards and surrounding landscape in the Montérégie administrative region of southern Quebec, Canada.

14

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

hive densities and all sites were within flight distance of an orchard that did rent honey bees. To control for between-cultivar variation in fruit and seed set we studied McIntosh apples, the leading cultivar in the northeastern United States and eastern Canada (Hampson and Kemp, 2003). Our sampling sites within each orchard were 100 m in length and 50 m in width (Appendix A, Supporting information), with placement largely determined by the availability of McIntosh trees. We measured a suite of covariates characterizing orchard management including tree basal trunk diameter (cm), tree density (trees m 2), as well as apple flower density and dandelion (Taraxacum officinale F.H. Wigg) abundance in the orchard understory. We monitored environmental factors that influence flowering and fruit development (Dennis et al., 2003) at twominute intervals within each site from May 5–19, 2012 using HOBO pendant temperature/light loggers and HOBO Pro v2 temperature/relative humidity loggers (Onset, Bourne, Massachusetts, USA). We sampled bees once at each site during the peak bloom of McIntosh apples (May 11–14, 2012) when study flowers were receptive, prior to petal falling and stigma browning (Losada and Herrero, 2013). Supplemental data were then gathered on bee diversity and behavior on later blooming varieties (May 15–18). Observation were made under clear to lightly overcast conditions from 9:00 to 18:00 PM Eastern Standard Time when ambient air temperatures were above 15  C and wind speeds <3.3 m s 1. Adapting the approach of Boyle-Makowski and Philogene (1985), we visually assigned bees to one of seven morphospecies: (1) honey bees, (2) bumble bees (Bombus spp. Latreille), (3) green metallic halictids (mostly Augochlora spp. (Say)), (4) large Andrena (Andrena carlini (Cockerell), Andrena duningii Cockerell, Andrena erythronii Robertson, Andrena milwaukeensis Graenicher, Andrena regularis Malloch, Andrena vicina Smith, Colletes spp. Say), (5) Osmia spp. Panzer, (6) small Andrena (all other Andrena) and (7) small black bees (Ceratina spp. Say Halictus spp. Latreille, Lasioglossum spp. Curtis). Colletes were grouped with large Andrena because of their morphological similarities. All morphospecies other than honey bees constituted the ‘wild bee’ group. Other pollinators (e.g., Syrphidae) in the orchards were not studied because of their marginal role in apple pollination in our region (Oliveira et al., 1980). We used three complementary and simultaneous sets of observations across orchards to characterize bee diversity and pollinator functional traits. Observations of bees visiting McIntosh apple trees, as opposed to herbs in the orchard understory, took place within 40 min time periods. The first observer examined functional traits limiting apple pollination following Vicens and Bosch (2000a), including: foraging rate (flowers min 1), foraging duration (average time (s) spent on flowers), the incidence of nectar versus pollen visits and whether or not the bee made contact with the stigma (i.e., incidence of ‘sideworking’). The amount of pollen deposited on stigmas per visit was not recorded given time constraints for the number of orchards and bee groups considered; moreover, apple pollination is less limited by the quantity of outcrossed pollen than by percent stigma contact (Vicens and Bosch, 2000a; Schneider et al., 2002; Sheffield et al., 2005; Sheffield, 2014). A second observer counted bees visiting apple blossoms in four evenly spaced clusters of McIntosh trees extending the length of the site (Appendix A, Supporting information), assigning each observation to morphospecies. Counts were partitioned evenly across a 6 min period per tree cluster. The time was noted every 1.5 min so that results could be cross-referenced with environmental conditions that may affect the activity of pollinating insects (Vicens and Bosch, 2000b). A third observer passed between the orchard rows with a sweep net to assess bee diversity at the site. Honey bees were

identified on the fly and tallied using a counter; wild bees were either caught with a sweep net or visually counted given their morphospecies identity. Netted specimens were frozen for species-level identification. Captured bees were identified to species using pertinent literature (Mitchell, 1960, 1962; Packer et al., 2007; Gibbs, 2010) as well as http://www.Discoverlife.org/ Specialists John Ascher, Bryan Danforth, and Jason Gibbs verified species identifications. Voucher specimens are deposited at the Lyman Entomological Museum, McGill University. We conducted an experiment on four evenly spaced McIntosh apple trees per site (Appendix A, Supporting information) to estimate the degree of pollen limitation and to correlate the incidence of fruit set of un-manipulated flowers with bee functional diversity. We studied only the centermost ‘king’ flower of the apple inflorescence, the least likely to be aborted if fruit is set (Dennis et al., 2003). On each study tree we tagged and then handpollinated 20 flowers with a single application of Red Delicious pollen (Firman Pollen Inc., Wakima, Washington, USA) and tagged 20 open-pollinated flowers freely visited by pollinators. Study flowers were exposed along the outer canopy of the trees, with branch height from the ground measured as a covariate (Dennis et al., 2003). Pollination services were quantified as the fruit set (May 23–25, 2012) and seed set (June 5–15, 2012) of apple blossoms. We asked orchardists not to apply chemical thinners to abort apples in the study area. Seed set was measured following Brault and Oliveira (1995) on eight pomes taken from a constant height throughout each of the study trees at each site. To determine an appropriate spatial scale at which to study landscapes surrounding our orchard sites, we estimated the average foraging range of the wild bee community. We calculated foraging range using an allometric conversion based on intertegular (IT) spans (Greenleaf et al., 2007) and then weighted these values by the abundance of each morphospecies across sites. The average foraging range was 680 m, which determines a 145 ha study landscape surrounding each sampling site. We focused on three land cover types to characterize the landscape surrounding sampled orchards: orchards, forests, and meadows. Orchards comprised all managed apple tree plantations in the 145 ha area, not only the studied orchard. Forests were primary or secondary growth, with closed hardwood canopies and understories of varying heights and densities. Meadows were open areas greater than 100 m2 dominated by grasses and forbs. We tallied the total area of each land cover type in the 145 ha foraging area around each site. We also calculated a ‘mean distance’ to forest and to meadow patches by averaging the Euclidean distances of vectors from site epicenters to all points of contact within meadows or along the forest edge in each study landscape. This ‘mean distance’ metric captures the spatial configuration of bee habitats surrounding each orchard (Fisher et al., 2009), gauging the mean length of foraging paths taken by bees relative to each orchard. The respective areas of the three land cover types estimate landscape composition and thus the availability of foraging and nesting resources surrounding a sampling site. We completed these landscape analyses using ArcGIS 10.1 (Environmental Systems Research Institute, Redlands, CA, USA). Land classes were interpreted from a composite of satellite and orthophoto imagery taken by the Ministère des ressources naturelles et de la faune du Québec (30 cm resolution) for the Montérégie region in 2009, DMTI Spatial (60 cm resolution) for Mont-St-Bruno in 2006 and Imex ltée (30 cm resolution) for MontSt-Hilaire in 2007. Interpretations were cross-referenced with data from the Quebec Forest Inventory Service and were verified in the field from June 22–25, 2012.

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

2.1. Data analysis All statistical analyses were conducted with R version 3.1.1. (R Core Team, 2014; R Project for Statistical Computing release 201407-11 http://www.r-project.org). 2.1.1. Assessing functional diversity Functional diversity of the bee community at each site was quantified using functional dispersion (Laliberté and Legendre, 2010). We chose FDis as a quantitative metric because it best combines the relative abundance of bees and the trait diversity elements of functional diversity (Laliberté and Legendre, 2010). The analysis measures the mean distance in multidimensional trait space of individual species to the centroid of all species. Functional dispersion was based on the relative abundance of morphospecies found across sites and a table describing multiple functional traits for each of these morphospecies. Functional traits of interest included aspects of bee behavior, environmentally mediated activity patterns, morphology and life history relevant to apple pollination. Metrics for bee behavior included apple foraging rate and the percentage of visits upon which each bee made contact with the stigma (Appendix B, Supporting information). Temporal and environmental preferences were synthesized as an ‘environmental index’ (Appendix B, Supporting information), with higher index values indicating greater bee activity during warmer, less humid conditions earlier in the day. Morphology was characterized in terms of bee size, as estimated from IT span measurements (Greenleaf et al., 2007). We also noted pollen carrying habit (Thorp, 2000) for each bee morphospecies, which refers to whether a bee carried pollen packed moist into corbiculae, and hence unavailable for pollination, or packed dry and hence available for pollination. Finally, two traits characterizing bee life history were taken from the literature: sociality (solitary or eusocial) and flight period (early: active from April–July; late: April–October) after Mitchell (1960, 1962),),Packer et al. (2007) and Gibbs (2010). To interpret our measure of functional dispersion, we visualized the multivariate trait space upon which it is based using principle coordinate analysis (PCoA). The ordination was performed from a Gower dissimilarity matrix of an n  p functional trait table (Laliberté and Legendre, 2010), where n represented ‘morphospecies’ and p ‘functional traits’. We then plotted a series of centroids for morphospecies scores, each weighted by the relative abundance of morphospecies found at a given orchard. This allowed for a visual assessment of changes in functional dispersion across orchards by examining the relative distance of morphospecies scores to the community-weighted centroid of each site. 2.1.2. Modeling ecosystem services We compared fruit set of hand- versus open-pollinated flowers to identify the degree of pollen limitation in the orchard system. We used a logistic generalized linear mixed model (GLMM; Bates et al., 2012) to predict the probability of fruit set as a function of pollination treatment (hand- versus open-pollinated), which was used as a fixed effect. We included trees nested within sites as random effects to account for the hierarchical structure of the sampling design. If hand-pollinated flowers were found to set fruit more often than open-pollinated flowers, this would indicate the orchards were pollen limited. Two separate GLMM analyses were used to study factors influencing the incidence of fruit set and the number of seeds produced per apples, respectively. In both analyses trees nested in sites were treated as random effects; we assumed fruit set followed a binomial error distribution and seed set a Poisson error distribution. Fixed effects were organized into four variable groups: bee FDis environmental variables, landscape metrics, and orchard management variables (Appendix C, Supporting

15

information). The environmental parameters were average values of temperature and light conditions over the period of flower development, pollination and early fruit development. Basal trunk diameters were specific to the trees studied for fruit and seed set. Branch heights were only available for the fruit set analysis as apples sampled for seed set were taken at a constant height. Variables with correlation coefficients greater than 0.5 were not included in the same model; meadow area could not be considered in these analyses because of collinearity with FDis. Explanatory variables were standardized to a mean of zero and a standard deviation of one to allow comparison of model parameter estimates. To study the relationship between bee functional diversity and the surrounding landscape, we used a multiple linear regression to model variation in FDis with landscape metrics as well as environmental and orchard management variables (Appendix C, Supporting information). In these analyses environmental covariates were averaged only over the period when bees were being sampled in each orchard. We checked for collinearity and standardized explanatory variables as in the analysis of fruit set and seed set. A model averaging approach was taken in model selection, which allowed us to study the uncertainty when quantifying the precision of a given coefficient (Johnson and Omland, 2004). We  , 2013) to used an automated model selection program (Barton generate all possible variable combination for the saturated model of each analysis. To restrict model complexity a maximum of five parameters were generated per permutation. We ranked models according to their AIC values calculated from maximum likelihood criteria, and estimated associated Akaike weights and relative  , 2013). Relative importance is calculatimportance scores (Barton ed as the sum of the Akaike weights over all of the models in which the parameter of interest appears (Johnson and Omland, 2004). We screened interactions of interest between variables within variable groups, but all had importance scores less than 0.60 and were dropped in further analyses. We then estimated model-averaged partial regression coefficients for each covariate as well as their 95% confidence intervals. Covariates were considered important if their summed Akaike weights were above 0.60 and significant if their confidence intervals did not include zero. We subjected the most parsimonious ‘top models’ (i.e., DAICc from the best model <2.0) to standard diagnostic testing. 3. Results 3.1. Field sampling results A total of 4686 bees were observed on McIntosh apple trees, 18% of which were wild bees and 82% honey bees. Wild bee abundance ranged from 1 to 55% across orchards, and honey bees 45–99%. A total of 36 wild bee species in eight genera and five families (Appendix D, Supporting information) were identified as having visited McIntosh trees. The most abundant and diverse genus of wild bee was Andrena (627 individuals and 17 species). Bumble bees (176 individuals) and small black bees (55 individuals) were less common. The green metallic halictids (three individuals) and Osmia (four individuals) were so seldom seen that neither group was considered in the morphospecies analyses. 3.2. Bee functional traits mediating pollination services The PCoA of the functional trait table (Table 1) explained a total of 92% of the variance along the first two axes of the ordination. Principle coordinate axis one (PC1; 79% variance) polarized honey bee and bumble bee traits against those of both Andrena groups. Traits associated with Andrena that are directly opposed to honey

16

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

bees and bumble bees, and hence are most correlated with PC1, include: early flight period, long foraging duration, dry pollen carrying habit, relatively small body size, and that they are solitary bees. Unlike Andrena, which nearly always made contact with the apple stigma, honey bees sideworked for 38% of all visits. Plotting the community-weighted centroid for each site in the trait space (Fig. 2) clearly indicates that sites with high FDis are not merely dominated by honey bees but are characterized by a greater proportion of wild bees, especially large and small Andrena. 3.3. Model averaging for pollination services and bee diversity Open-pollinated flowers set fruit half as often (0.54 fewer times) as hand-pollinated flowers (P < 0.001), indicating pollen limitation is a significant factor determining fruit set and seed set of open-pollinated flowers across sites. Considering patterns of bee diversity (Table 2), it is clear that in this pollen limited system functional dispersion is an important predictor of fruit set for open pollinated flowers (w = 0.68, b = 0.38  0.37CI). Since an additional test regressing fruit set in hand-pollinated flowers on FDis was not significant (b = 1.90, P > 0.10), pollen limitation was not confounded by factors such as nutrient limitation or plant growth. Seed set of open pollinated flowers was also positively determined by FDis (w = 0.79, b = 0.08  0.07CI), and negatively predicted by distance from meadow (w = 0.98, b = 0.12  0.06CI) and by total orchard area (w = 0.76, b = 0.07  0.06CI). Model averaging (Table 3) indicated that FDis is most positively related to total meadow area (w = 1.00, b = 0.11  0.05CI) and total forest area (w = 0.86, b = 0.05  0.04CI). All remaining variables were considered non-significant as they had relative importance scores below 0.60 and confidence intervals associated with the partial coefficients that overlapped zero. Spatial autocorrelation in the residuals of all models was examined using semi-variograms. We used the most parsimonious models in the ‘top model’ set (i.e., DAIC from the best model <2.0) to generate land use recommendations enabling increased pollination services (Fig. 3). Fruit set was modeled by FDis (b = 0.35, P < 0.05). Seed set was a function of FDis (b = 0.09, P < 0.01) and mean distance from meadow (b = 0.11, P < 0.001). The FDis linear model (R2adj = 0.56, P < 0.001) had total meadow area (b = 0.09, P < 0.001) as an explanatory variable. A level of seed set indicative of high fruit quality (Brault and Oliveira, 1995) is associated with a mean

Fig. 2. Principle coordinate analysis biplot of the morphospecies trait table. Diamonds represent morphospecies (BB: bumble bees; HB: honey bees; LA: large Andrena; SA: small Andrena; SB: small black bees) and circles the communityweighted centroids of species coordinates at each of the 20 sites. The size of each centroid is scaled to its respective value of FDis; the centroid with the highest value of FDis in black for emphasis. The size of each morphospecies diamond is scaled to its respective relative abundance at the site with the highest value of FDis.

distance to surrounding meadow of 400 m, FDis of 0.25, and in turn a total meadow area of 7 ha (5% of the surrounding landscape). 4. Discussion The fundamental challenge to apple pollination is the transfer of compatible pollen to flowers during the brief period of floral receptivity in spring. Despite honey bees having been maintained at recommended densities, our supplemental pollination experiment indicated that the orchards studied were generally pollen limited. With the proviso that the efficacy of our hand pollination of experimental flowers exceeds that expected from natural pollinators (Aizen and Harder, 2007), our results are extendible to similar pollen-limited systems. In our orchards, the increased fruit set and seed set of unmanipulated flowers is clear evidence that the functional diversity of the bee guild is important for overcoming the apple pollination deficit. Wild bees have functional traits that are distinct from and complementary to those of honey bees. The availability of wild bee pollinators in orchards depends on forest and meadow habitats in the surrounding landscape that provide foraging and nesting resources before and after apples bloom, respectively. Apple pollination services are thus contingent on the complementarity in functional traits between managed and unmanaged bees, as well as complementary habitat use by bees in the landscapes surrounding orchards. 4.1. Complementarity in bee functional traits enhance pollination services Functional diversity can improve pollination services through resource use complementarity, which arises in the relative abundance of species in a community and the degree to which species-specific traits influencing pollination vary in the community. Honey bees, the dominant bee species in the studied apple orchards, exhibit many traits that facilitate apple pollination: they are easily transportable, loyal to the crops they visit and can be maintained at high densities at the onset of the season when other pollinators are less abundant (Free, 1993). The prevalence of honey bee sideworking that we observed, however, suggests that orchards dominated only by honey bees are less effectively pollinated (Robinson and Fell, 1981; Schneider et al., 2002). The means by which functional diversity promotes apple pollination will depend on how functional traits of wild bees compensate for honey bee inadequacies. Andrena was the most abundant wild bee genera in our orchards and andrenid bees exhibit multiple functional traits that complement honey bee pollination of apples. Pollen adhering to the dense and specialized pubescence of Andrena species (Gardner and Ascher, 2006) is readily transferred to the stigma (Thorp, 2000), much in contrast to pollen packed moist into the hind leg scopae of honey bees (Westerkamp, 1991). Discounting the poorly transferable corbicular pollen pellets in honey bees, andrenid bees carry more apple pollen on their bodies (Kendall and Solomon, 1973) and deposit twice as much pollen per visit (M. Park, pers. comm. 2014). Both andrenid groups nearly always made contact with the stigma when visiting the apple blooms and had long foraging durations, exhibiting high pollinator efficacy per visit (Thomson, 1986; Vicens and Bosch, 2000a). The slow foraging rate and dependence on agreeable weather (Appendix B, Supporting information) of andrenids may, however, limit their capacity to provide sufficient pollination services during the brief period of apple stigma receptivity. Bumble bees have functional traits that offset deficiencies in both honey bees and andrenid apple pollination. Consistent with observations by Jacob-Remacie (1989), bumble bees in our orchards never sideworked and had the highest floral visitation

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

17

Table 1 The functional traits for bee morphospecies regarding: mean (SD, n) foraging duration (s), foraging rate (flower min 1) and inter-tegulare (IT) span (mm); percentage of visits where contact is made with stigma; environmental index scores (higher values indicate preference for warmer, less humid conditions earlier in the day); if pollen in corbiculae is moistened or left dry; flight period (early: active from April–July; late: April–October); and if the bee was solitary or social. Different superscript letters indicate significant differences (Kruskall–Wallis Test; P < 0.05).

Foraging duration (s) Foraging rate (flr min 1) IT span (mm) Stigma contact (%) Environmental index Pollen carrying habit Flight period Sociality a

Bumble bee

Honey bee

L. Andrena

S. Andrena

S. black bee

2.9a (1.3, 43)

8.7b (7.0, 56)

23.4c (35.6, 52)

20.9cd (18.7,47)

24.9d (19.7, 41)

14.3a (5.3, 43)

6.4b (3.2, 56)

4.4c (3.2, 52)

3.6c (2.1,47)

3.2c (2.0, 41)

5.5a (0.5, 28) 1 0.4 Moist Late Social

3.2b (0.1, 20) 0.62 0.7 Moist Late Social

2.7c (0.4, 38) 1 1.1 Dry Early Solitary

2.2d (0.3, 53) 0.97 0.9 Dry Early Solitary

1.6e (0.2, 32) 0.95 0.1 Dry Late Sociala

Although the “small black bee” group includes both solitary and social bee species, all species captured in this study were social.

Table 2 Estimated coefficients (b), their 95% confidence intervals and importance values (w) are given per model parameter for fruit set and seed set model averaging of GLMM analyses. Significant terms with confidence intervals not overlapping with zero are in bold. Fruit set

Seed set

b

w

Lower CI

Upper CI

w

b

Lower CI

Upper CI

Functional dispersion

0.68

0.38

0.01

0.76

0.79

0.08

0.01

0.15

Landscape metrics Mean meadow distance (m) Total orchard area (m2) Mean forest distance (m) Total forest area (m2)

0.34 0.25 0.32 0.29

-0.18 -0.04 -0.17 0.13

0.57 0.43 0.53 0.25

0.20 0.35 0.19 0.51

0.98 0.76 0.22 0.28

0.12 0.07 0.01 0.03

0.18 0.13 0.05 0.04

0.06 0.01 0.08 0.09

Orchard management Trunk diameter (cm) Branch height (m) Trees density (trees m

0.03 0.13 0.16

0.21 0.12 0.25

0.28 0.38 0.57

0.34

0.03

0.08

0.02

)

0.25 0.35 0.32

0.28

0.03

0.05

0.11

Orchard environment Light intensity (lx) Average temperature ( C)

0.28 0.29

0.07 0.13

0.38 0.48

0.52 0.23

0.26 0.20

0.00 0.01

0.09 0.06

0.09 0.05

2

rates. Thomson and Goodell (2001) reported that bumble bees deposited more pollen grains on apple stigmas than sideworking honey bees, and Garratt et al. (2013) found bumble bee visitation to be important for apple seed set. Since bumble bees can be active in temporal and environmental conditions unfavorable for both honey bees and andrenids (Appendix B, Supporting information) Table 3 Estimated coefficients (b), their 95% confidence intervals and importance values (w) are given per model parameter for the bee functional dispersion model averaging of simple linear model analysis. Significant terms with confidence intervals not overlapping with zero are in bold. Bee functional dispersion w Landscape metrics Total meadow area (m2) Total forest area (m2) Mean forest distance (m) Total orchard area (m2) Mean meadow distance (m)

1.00 0.86 0.21 0.10 0.10

b

Lower CI 0.11 0.05 0.02 0.01 0.00

0.06 0.01 0.07 0.04 0.04

Upper CI 0.16 0.09 0.02 0.05 0.05

they can increase the potential for pollination of apples despite the vagaries of spring weather (Free, 1993). This potential, however, is offset by their relatively low abundance in the wild bee community; the bumble bees when apples are in bloom are queens emerging from winter dormancy and have yet to establish hives with abundant workers (Gardner and Ascher, 2006). To summarize, honey bees, andrenids and bumble bees have complementary traits associated with foraging and resource use that in combination are important in successfully pollinating apples. Honey bees were abundant but less effective pollinators, Andrena efficacious and abundant but slow foragers, and bumble bees fast and able to forage in unfavorable weather but in low numbers. We showed that functional dispersion in the bee community positively predicted fruit set and seed set in our orchards. Pollination services were greater in orchards with a mix of bees exhibiting different, complementary traits as opposed to orchards that were more dependent on honey bees alone. 4.2. Complementary habitats promote bee functional diversity

Orchard management Trunk diameter (cm) Trees density (trees m 2) Dandelions abundance

0.46 0.16 0.11

0.04 0.02 0.01

0.01 0.06 0.04

0.08 0.02 0.05

Orchard environment Temperature ( C) Light intensity (lx) Wind rank

0.19 0.12 0.11

0.02 0.01 0.00

0.02 0.06 0.05

0.07 0.03 0.05

A key question is what can be done to increase the abundance of wild bees and thus functional diversity of the bee guild securing pollination services in apple orchards. This is a question not simply of orchard practice but also of managing the landscape surrounding an orchard. While honey bee hives can readily be brought to an orchard when needed, increasing the abundance of wild bees requires providing suitable foraging resources and nesting sites in

18

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

Fig. 3. Panel (A) probability of apple fruit set versus functional dispersion (FDis) GLMM; panels (B) and (C) added-variable plots for apple seed set versus FDis and mean distance to meadow (m) GLMM; panel (D) FDis versus meadow area (m2) simple linear model. Dashed lines indicate 95% confidence envelopes. An increase in FDis indicates an increase in the relative abundance of different bee morphospecies exhibiting different functional traits associated with pollination efficacy.

the surrounding landscapes prior to and following the apple bloom. During springtime the understory herbs and shrubs in forests of eastern North America are important pollen and nectar resources for the primary wild bee groups in orchards: the andrenids and bumble bee queens (Taki et al., 2007; Watson et al., 2011). The closing of the forest canopy by late May coincides with the end of apple blossoming in our region, at which point the shaded forest understory also becomes less favorable for flowering and bee foraging activity (Mandelik et al., 2012). Meadows, which have many flowering species later in summer, provide a natural complement to forests (Mandelik et al., 2012) and have been found to support the greatest diversity of wild bees when compared with commercial apple orchards (Sheffield et al., 2013). Since current bee populations derive from resources available to the prior generation (Roulston and Goodell, 2011), floral resources in forests and meadows one year influence bee abundance in orchards the following year. Therefore, it is necessary to consider the temporal and spatial variability in bee foraging resources within and across generations to promote pollination services by wild bees in orchards. Although the availability and quality of surrounding habitats mediate the supply of wild bee pollinators, the total area of orchard in the landscape dictates the demand for pollinators. Large

expanses of blooming apple trees can simply exceed the capacity of wild bees to provide adequate pollination services (Holzschuh et al., 2011). We found that as the total area of surrounding orchards increased, seed set in our study orchards decreased. Marini et al. (2012) likewise found that native bee abundance and diversity was low in orchard-dominated landscapes, because of the increased distance to orchard interior from peripheral bee habitats, seasonal pesticide and fungicide applications, and lack of floral resources when apples are not in bloom. There clearly is a balance to be struck between increasing the area in apple production and enhancing wild bee habitat and thus pollination services per unit area of orchard; a compromise may be to increase floral species richness in the orchard understory, which has been found to favor bee faunal assemblages (García and Miñarro, 2014). Our results can be used to generate management objectives for our study region and comparable temperate settings that target the provision of wild bee diversity and pollination services. Given that greater forest area is difficult to achieve in the short term, we recommend the greater integration of open field habitats in regional agroecosystems. Meadow and grassland restoration have been successfully adopted for bee conservation in Europe (Albrecht et al., 2007). These sorts of land stewardship programs for pollinator refugia require coordinated efforts among farmers to avoid negative externalities and common pool resource problems.

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

5. Conclusions We found that a mix of managed and unmanaged bees exhibiting complementary functional traits ensured high apple fruit set and seed set. We are not recommending that orchardists abandon the use of honey bees in apple pollination in favor of wild bees. Rather, orchardists and land use managers may consider the spatio-temporal complementarity of wild bee habitat surrounding orchards and seek opportunities to manage this habitat to enhance pollination services. Acknowledgments We gratefully acknowledge the Natural Science and Engineering Research Council of Canada, the Fonds de Recherche Nature et Technologies Québec, the Quebec Centre for Biodiversity Sciences and the Fédération des Producteurs de Pommes du Québec for funding and support. AG is supported by the Canada Research Chair program. We thank the apple growers in the Montérégie region for their willingness to provide access to their orchards, Julien MasséJodoin, Natasha Salter and Sarah Saldahna for field assistance, John Ascher, Bryan Danforth and Jason Gibbs for their help with insect identification, and James Thomson for comments on the manuscript. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.agee.2014.10.018. References Aizen, M.A., Harder, L.D., 2007. Expanding the limits of the pollen-limitation concept: effects of pollen quantity and quality. Ecology 88, 271–281. Aizen, M.A., Harder, L.D., 2009. The global stock of domesticated honey bees is growing slower than agricultural demand for pollination. Curr. Biol. 19, 915–918. Albrecht, M., Duelli, P., Müller, C., Kleijn, D., Schmid, B., 2007. The Swiss agrienvironment scheme enhances pollinator diversity and plant reproductive success in nearby intensively managed farmland. J. Appl. Ecol. 44, 813–822.  K., 2013. MuMIn: Multi-model inference. R package version 1.9.0., URL Barton http://CRAN.R-project.org/package=MuMIn. Bates D., Maechler M., Bolker B., 2012. lme4: Linear mixed-effects models using S4 classes. R package version 0.999999-0, URL http://CRAN.R-project.org/ package=lme4. Blüthgen, N., Klein, A.-M., 2011. Functional complementarity and specialisation: the role of biodiversity in plant–pollinator interactions. Basic Appl. Ecol. 12, 282–291. Boyle-Makowski, R., Philogene, B., 1985. Pollinator activity and abiotic factors in an apple orchar. Can. Entomol. 117, 1509–1521. Brault, A.-M., Oliveira, D.D., 1995. Seed number and an asymmetry index of mcintosh apples. HortScience 30, 44–46. Dennis, F., Ferree, D., Warrington, I., 2003. Flowering, pollination and fruit set and development. In: Ferree, D., Warrington, I. (Eds.), Apples: Botany, Production and Uses. CABI Publishing, Cambrige, pp. 153–166. Fisher, B., Turner, R.K., Morling, P., 2009. Defining and classifying ecosystem services for decision making. Ecol. Econ. 68, 643–653. Free, J.B., 1993. Insect Pollination of Crops. Academic Press, London. Fründ, J., Dormann, C.F., Holzschuh, A., Tscharntke, T., 2013. Bee diversity effects on pollination depend on functional complementarity and niche shifts. Ecology 94, 2042–2054. García, R.R., Miñarro, M., 2014. Role of floral resources in the conservation of pollinator communities in cider-apple orchards. Agri. Ecosyst. Environ. 183, 118–126. Gardner, K., Ascher, J., 2006. Notes on the native bee pollinators in New York apple orchards. J. N.Y. Entomol. Soc. 114, 86–91. Garibaldi, L.A., Steffan-Dewenter, I., Winfree, R., Aizen, M.A., Bommarco, R., Cunningham, S.A., Kremen, C., Carvalheiro, L.G., Harder, L.D., Afik, O., Bartomeus, I., Benjamin, F., Boreux, V., Cariveau, D., Chacoff, N.P., Dudenhöffer, J.H., Freitas, B. M., Ghazoul, J., Greenleaf, S., Hipólito, J., Holzschuh, A., Howlett, B., Isaacs, R., Javorek, S.K., Kennedy, C.M., Krewenka, K.M., Krishnan, S., Mandelik, Y., Mayfield, M.M., Motzke, I., Munyuli, T., Nault, B.A., Otieno, M., Petersen, J., Pisanty, G., Potts, S.G., Rader, R., Ricketts, T.H., Rundlöf, M., Seymour, C.L., Schüepp, C., Szentgyörgyi, H., Taki, H., Tscharntke, T., Vergara, C.H., Viana, B.F., Wanger, T.C., Westphal, C., Williams, N., Klein, A.M., 2013. Wild pollinators enhance fruit set of crops regardless of honey bee abundance. Science 339, 1608–1611.

19

Garratt, M.P.D., Truslove, C.L., Coston, D.J., Evans, R.L., Moss, E.D., Dodson, C., Jenner, N., Biesmeijer, J.C., Potts, S.G., 2013. Pollinator deficits in UK apple orchards. J. Pollinat. Ecol. 11, 1–6. Gibbs, J., 2010. Revision of the metallic species of Lasioglossum (Dialictus) in Canada (Hymenoptera, Halictidae, Halictini). Zootaxa 2591, 1–382. Greenleaf, S.S., Williams, N.M., Winfree, R., Kremen, C., 2007. Bee foraging ranges and their relationship to body size. Oecologia 153, 589–596. Hampson, C.R., Kemp, H., 2003. Characteristics of important apple cultivars. In: Ferree, D.C., Warrington, I.J. (Eds.), Apples: Botany, Production, and Uses. CABI Publishing, Cambrige, pp. 61–90. Hoehn, P., Tscharntke, T., Tylianakis, J.M., Steffan-Dewenter, I., 2008. Functional group diversity of bee pollinators increases crop yield. Proc. R. Soc. B: Biol. Sci. 275, 2283–2291. Holzschuh, A., Dormann, C.F., Tscharntke, T., Steffan-Dewenter, I., 2011. Expansion of mass-flowering crops leads to transient pollinator dilution and reduced wild plant pollination. Proc. R. Soc. B: Biol. Sci. 278, 3444–3451. Jacob-Remacie, A., 1989. Comportement de butinage de l’abeille domestique et des abeilles sauvages dans des vergers de pommiers en Belgique. Apidologie 20, 271–285. Johnson, J.B., Omland, K.S., 2004. Model selection in ecology and evolution. Trends Ecol. Evol. 19, 101–108. Kendall, D., 1973. The viability and compatibility of pollen on insects visiting apple blossom. J. Appl. Ecol. 10, 847–853. Kendall, D., Solomon, M., 1973. Quantities of pollen on the bodies of insects visiting apple blossom. J. Appl. Ecol. 10, 627–634. Kennedy, C.M., Lonsdorf, E., Neel, M.C., Williams, N.M., Ricketts, T.H., Winfree, R., Bommarco, R., Brittain, C., Burley, A.L., Cariveau, D., Carvalheiro, L.G., Chacoff, N. P., Cunningham, S.A., Danforth, B.N., Dudenhöffer, J.-H., Elle, E., Gaines, H.R., Garibaldi, L.A., Gratton, C., Holzschuh, A., Isaacs, R., Javorek, S.K., Jha, S., Klein, A. M., Krewenka, K., Mandelik, Y., Mayfield, M.M., Morandin, L., Neame, L.A., Otieno, M., Park, M., Potts, S.G., Rundlöf, M., Saez, A., Steffan-Dewenter, I., Taki, H., Viana, B.F., Westphal, C., Wilson, J.K., Greenleaf, S.S., Kremen, C., 2013. A global quantitative synthesis of local and landscape effects on wild bee pollinators in agroecosystems. Ecol. Lett. 16, 584–599. Klein, A.-M., Cunningham, S.A., Bos, M., Steffan-Dewenter, I., 2008. Advances in pollination ecology from tropical plantation crops. Ecology 89, 935–943. Kremen, C., Williams, N.M., Aizen, M.A., Gemmill-Herren, B., LeBuhn, G., Minckley, R., Packer, L., Potts, S.G., Steffan-Dewenter, I., Vazquez, D.P., 2007. Pollination and other ecosystem services produced by mobile organisms: a conceptual framework for the effects of land-use change. Ecol. Lett. 10, 299–314. Laliberté, E., Legendre, P., 2010. A distance-based framework for measuring functional diversity from multiple traits. Ecology 91, 299–305. Losada, J.M., Herrero, M., 2013. Flower strategy and stigma performance in the apple inflorescence. Scientia Horticulturae 150, 283–289. Mandelik, Y., Winfree, R., Neeson, T., Kremen, C., 2012. Complementary habitat use by wild bees in agro-natural landscapes. Ecol. Appl. 22, 1535–1546. Marini, L., Quaranta, M., Fontana, P., Biesmeijer, J.C., Bommarco, R., 2012. Landscape context and elevation affect pollinator communities in intensive apple orchards. Basic Appl. Ecol. 13, 681–689. Mitchell, M.G., Bennett, E.M., Gonzalez, A., 2013. Linking landscape connectivity and ecosystem service provision: current knowledge and research gaps. Ecosystems 16, 894–908. Mitchell, T.B., 1960. Bees of The Eastern United States. North Carolina Agricultural Experiment Station, North Carolina. Mitchell, T.B., 1962. Bees of The eastern United States. North Carolina Agricultural Experiment Station, North Carolina. Neumann, P., Carreck, N.L., 2010. Honey bee colony losses. J. Apicult. Res. 49, 1–6. Oliveira, D.D., Pion, S., Paradis, R.O., 1980. La pollinisation et la production de pommes. La Phytoprotection des Vergers de Pommiers au Québec. Agri. Canada 51–61 St-Jean-sur-Richelieu. Packer, L., Genaro, J.A., Sheffield, C.S., 2007. The bee genera of eastern Canada. Can. J. Arthropod Identif. 3, 1–32. R Core Team, 2014. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/. Rader, R., Reilly, J., Bartomeus, I., Winfree, R., 2013. Native bees buffer the negative impact of climate warming on honey bee pollination of watermelon crops. Glob. Change Biol. 19, 3103–3110. Robinson, W., Fell, R., 1981. Effect of honey bee foraging behaviour on delicious apple set. HortScience 16, 326–328. Roulston, T., Goodell, K., 2011. The role of resources and risks in regulating wild bee populations. Ann. Rev. Entomol. 56, 293–312. Schneider, D., Stern, R.A., Eisikowitch, D., Goldway, M., 2002. The relationship between floral structure and honeybee pollination efficiency in ‘Jonathan’ and ‘Topred’ apple cultivars. J. Hortic. Sci. Biotechnol. 77, 48–51. Sheffield, C.S., 2014. Pollination, seed set and fruit quality in apple: studies with Osmia lignaria (Hymenoptera: Megachilidae) in the Annapolis Valley, Nova Scotia, Canada. J. Pollinat. Ecol. 12, 120–128. Sheffield, C.S., Kevan, P.G., Pindar, A., Packer, L., 2013. Bee (Hymenoptera: Apoidea) diversity within apple orchards and old fields in the Annapolis Valley, Nova Scotia, Canada. Can. Entomol. 145, 94–114. Sheffield, C.S., Smith, R.F., Kevan, P.G., 2005. Perfect syncarpy in apple (Malus x domestica ‘Summerland McIntosh’) and its implications for pollination, seed distribution and fruit production (Rosaceae: Maloideae). Ann. Bot. 95, 583–591. Taki, H., Kevan, P.G., Ascher, J.S., 2007. Landscape effects of forest loss in a pollination system. Landscape Ecol. 22, 1575–1587.

20

K.T. Martins et al. / Agriculture, Ecosystems and Environment 200 (2015) 12–20

Thomson, J.D., 1986. Pollen transport and deposition by bumble bees in Erythronium: influences of floral nectar and bee grooming. J. Ecol. 74, 329–341. Thomson, J.D., Goodell, K., 2001. Pollen removal and deposition by honeybee and bumblebee visitors to apple and almond flowers. J. Appl. Ecol. 38, 1032–1044. Thorp, R.W., 2000. The collection of pollen by bees. Pollen and Pollination. Springer, pp. 211–223. Vicens, N., Bosch, J., 2000a. Pollinating efficacy of Osmia cornuta and Apis mellifera (Hymenoptera: Megachilidae, Apidae) on ‘red delicious’ apple. Environ. Entomol. 29, 235–240.

Vicens, N., Bosch, J., 2000b. Weather-dependent pollinator activity in an apple orchard, with special reference to Osmia cornuta and Apis mellifera (Hymenoptera: Megachilidae and Apidae). Environ. Entomol. 29, 413–420. Watson, J., Wolf, A., Ascher, J., 2011. Forested landscapes promote richness and abundance of native bees (Hymenoptera: Apoidea: Anthophila) in Wisconsin apple orchards. Environ. Entomol. 40, 621–632. Westerkamp, C., 1991. Honeybees are poor pollinators—why? Plant Syst. Evol. 177, 71–75.