-
1158 | wileyonlinelibrary.com/journal/jane J Anim Ecol.
2019;88:1158–1167.© 2019 The Authors. Journal of Animal Ecology ©
2019 British Ecological Society
1 | INTRODUCTION
Pollinators’ interactions with plants tend to be generalized
when viewed at the species level, with most pollinator species
collect-ing floral resources from multiple plant families (Waser,
Chittka, Price, Williams, & Ollerton, 1996). Although such a
species-level understanding of plant–pollinator interactions yields
important insights about communities, for example with respect to
their sta-bility (Bascompte, Jordano, & Olesen, 2006) and
restoration (Kaiser-Bunbury et al., 2017), it also obscures the
biology of individual-level interactions, which can strongly affect
patterns of pollen transfer
between plants (Brosi, 2016). Foraging bout specialization, or
the specialization of an individual pollinator within a foraging
bout (i.e. a single foraging trip), affects plant reproductive
success because individuals that move sequentially between plants
of the same spe-cies transfer more conspecific pollen and less
heterospecific pollen between flowers (Campbell & Motten, 1985;
Morales & Traveset, 2008). Here, we use the term foraging bout
specialization to refer to patterns of both random foraging (e.g.
when an individual happens to specialize, because there is only one
rewarding plant species in its foraging path) and non-random
foraging (i.e. floral constancy, when an individual specializes
despite there being multiple rewarding
Received: 5 November 2018 | Accepted: 2 April 2019DOI:
10.1111/1365-2656.13003
R E S E A R C H A R T I C L E
Specialist foragers in forest bee communities are small, social
or emerge early
Colleen Smith1,2 | Lucia Weinman1,2 | Jason
Gibbs3 | Rachael Winfree2
1Graduate Program in Ecology & Evolution, Rutgers
University, New Brunswick, New Jersey2Department of Ecology,
Evolution, and Natural Resources, Rutgers University, New
Brunswick, New Jersey3Department of Entomology, University of
Manitoba, Winnipeg, Manitoba, Canada
CorrespondenceColleen SmithEmail: [email protected]
Funding informationXerces Society for Invertebrate
Conservation; Natural Resources Conservation Service; Garden Club
of America
Handling Editor: Julian Resasco
Abstract1. Individual pollinators that specialize on one plant
species within a foraging bout
transfer more conspecific and less heterospecific pollen,
positively affecting plant reproduction. However, we know much less
about pollinator specialization at the scale of a foraging bout
compared to specialization by pollinator species.
2. In this study, we measured the diversity of pollen carried by
individual bees forag-ing in forest plant communities in the
mid-Atlantic United States.
3. We found that individuals frequently carried low-diversity
pollen loads, suggest-ing that specialization at the scale of the
foraging bout is common. Individuals of solitary bee species
carried higher diversity pollen loads than did individuals of
social bee species; the latter have been better studied with
respect to foraging bout specialization, but account for a small
minority of the world’s bee species. Bee body size was positively
correlated with pollen load diversity, and individuals of
polylectic (but not oligolectic) species carried increasingly
diverse pollen loads as the season progressed, likely reflecting an
increase in the diversity of flowers in bloom. Furthermore, the
seasonal increase in pollen load diversity was stronger for bees
visiting trees and shrubs than for bees visiting herbaceous
plants.
4. Overall, our results showed that both plant and pollinator
species’ traits as well as community-level patterns of flowering
phenology are likely to be important determinants of
individual-level interactions in plant–pollinator communities.
K E Y WO RD S
floral constancy, foraging behaviour, individuals, phenology,
pollinator
-
| 1159Journal of Animal EcologySMITH eT al.
plant species in its foraging path), both of which have similar
con-sequences for plants in terms of their effects on pollen
transfer. Despite the potential functional importance of pollinator
foraging bout specialization, little is known about its prevalence
across a wide diversity of pollinator species.
Most of what we know about foraging bout specialization comes
from studies of bees in only two of the roughly 440 bee genera:
Apis (honeybees) and Bombus (bumblebees). Studies of bees in these
genera suggest that foraging bout specialization is quite common.
Foraging honeybees have been found to carry pure pollen loads an
overwhelming majority (i.e. >90%) of the time, and bumblebees
also tend to carry pure pollen loads more often than not, though
less fre-quently than honeybees (range 49% to 76% of the time;
Free, 1963; Grant, 1950; Leonhardt & Blüthgen, 2012). However,
studies of hon-eybees and bumblebees might tell us little about the
prevalence of foraging bout specialization in other bee taxa, for
two reasons. First, both Apis and Bombus are eusocial, while 90% of
bee species are not (Danforth, Cardinal, Praz, Almeida, &
Michez, 2013). Second, and more generally, if bee species vary in
their foraging behaviour, then studies of just a few bee species
might not be very informative for understanding foraging bout
specialization across many bee species. In the few pollen load
studies of bees from other genera, we see that, in some species, as
many as 97% of individuals carry pure pollen loads while foraging
(in stingless bees; Ramalho, Giannini, Malagodi-Braga, &
Imperatriz-Fonseca, 1994), but in other species, only 34% of
individuals carry pure pollen loads (in mason bees; Eckhardt,
Haider, Dorn, & Müller, 2014). However, because research has
focused on Apis and Bombus, we know very little about how common
foraging bout specialization is in bees across a wider diversity of
genera and life histories (but see Grant, 1950; Motten, 1986;
Pornon, Andalo, Burrus, & Escaravage, 2017).
The life-history traits of pollinators, and also of the plant
spe-cies in their environment, could affect how commonly
individuals exhibit foraging bout specialization. First, generalist
solitary bee species might be less specialized in their foraging
behaviour than social bee species. Social species may have a
tendency to specialize while foraging for a number of reasons
related to cognitive ability and resource partitioning (Chittka,
Thomson, & Waser, 1999; Jones & Agrawal, 2017), and two
experiments have shown that individuals of social bees species
exhibited foraging bout specialization more often than individuals
of solitary bee species; in both studies, the social bees were
faster to learn to associate floral rewards with a plant species’
morphology (Amaya-Márgez & Wells, 2008; Dukas & Real,
1991). However, neither study included more than one social or
solitary bee species, and there has yet to be a rigorous test of
the pattern that social bees are more specialized foragers than
solitary ones, where individuals are compared across multiple
social and sol-itary bee species foraging in the same
environment.
Next, bee body size, bee dietary breadth and the growth form of
the available plants could each affect how specialized a forager
will be, because they affect the number of rewarding plant species
a foraging bee encounters. First, we expect individuals of large
bee species to carry more diverse pollen loads, even if this is by
chance
alone. Large species require more overall pollen for their
offspring and can fit a larger volume of pollen in their pollen
loads (Müller et al., 2006; Ramalho et al., 1994). They will thus
likely visit more individual flowers during a single foraging bout,
which would cause them to also encounter more plant species.
Second, individuals of generalist bee species (hereafter
“polylectic” bees) are able to use pollen from more plant species
than individuals of specialist bee spe-cies (hereafter
“oligolectic” bees), which are physiologically and be-haviourally
limited to only a few plant taxa for pollen (Praz, Müller, &
Dorn, 2008; Williams, 2003). Thus, we expect individuals of
polylec-tic species to carry more diverse pollen loads than
individuals of oligolectic species (Cripps & Rust, 1989;
Müller, 1996). Finally, we expect bees that visit herbs to carry
more diverse pollen loads than bees that visit trees or shrubs,
because the large floral displays pro-duced by trees and shrubs
could induce a foraging bee to stay on one plant longer and save
the search costs of foraging elsewhere. This should lead to fewer
encounters with other plant species (and also an increase in the
rate of selfing, a mating cost from the plant's per-spective;
Kunin, 1993, Barrett & Harder, 1996, Castilla et al., 2017).
Alternatively, each of these traits may matter very little if most
bees exhibit floral constancy, a behaviour that increases foraging
effi-ciency by removing the cognitive demands associated with
switching between plant species (Chittka et al., 1999; Gegear &
Laverty, 2005). Because all bees benefit from increased foraging
efficiency, floral constancy might be very common, and there might
be little variation between individuals for these traits to
explain.
Finally, foraging bout specialization by pollinators could vary
over time with seasonal changes in the diversity of plant species
in bloom. All regions exhibit seasonal changes in the diversity of
floral resources available to pollinators (Ogilvie & Forrest,
2017). In regions with shortened growing seasons, floral diversity
tends to exhibit uni-modal or bimodal patterns, with diversity
increasing as environmen-tal conditions become more favourable
(e.g. Anderson & Hubricht, 1940, Heinrich, 1976, Kochmer &
Handel, 1986, Morales, Dodge, & Inouye, 2005, Makrodimos,
Blionis, Krigas, & Vokou, 2008, Craine, Wolkovich, Towne, &
Kembel, 2012). Studies show that individuals respond to increases
in floral diversity by visiting more plant species within a
foraging bout (Flanagan, Mitchell, & Karron, 2011; Kunin, 1993;
Lucas et al., 2018). However, if some bees (e.g. social bees, small
bees, oligolectic bees) have a strong tendency to exhibit for-aging
bout specialization, changes in floral diversity over time might
have little effect on foraging bout specialization in these
species.
In this study, we measure foraging bout specialization for
for-est bee communities native to eastern US temperate deciduous
forests. These forests, which once covered much of eastern North
America, support diverse native bee communities in which adult bees
of many species are active for two to three months, from the start
of spring in March through the closing of the forest canopy in May
(Harrison, Gibbs, & Winfree, 2018). Forest bees are
under-studied with respect to floral resource use and are largely
neglected by common pollinator conservation efforts, such as
pollinator floral restorations, which tend to only include
summer-blooming plants (Hicks et al., 2016; Wood et al., 2018);
thus, there is a great research
-
1160 | Journal of Animal Ecology SMITH eT al.
need for understanding the foraging needs of these species. We
also chose this system because it provides a natural gradient in
flower-ing diversity; only a few species flower in the earliest
weeks, but diversity steadily increases over the course of the
spring (Anderson & Hubricht, 1940; Heinrich, 1976; Kochmer
& Handel, 1986). We measured pollinator foraging patterns in
plant–pollinator communi-ties throughout this period by removing
and identifying the pollen collected into individual bees’ scopae
and corbiculae (pollen-car-rying structures). These pollen loads
provide a history of the plant species a female bee has visited for
pollen during a single foraging bout, because the bee removes the
pollen from her pollen-carrying structures when she returns to her
nest after each foraging bout. Pollen loads thus enable the
measurement of pollen foraging bout specialization across many
individuals and species.
To better understand the dynamics of pollen foraging across many
bee species, we ask the following: (a) What is the distribution of
pollen load diversities across individual bees collected while
for-aging? (b) Do species-level traits (bee sociality, bee body
size, bee dietary breadth or plant growth form) affect the
diversity of pollen carried by individuals? and (c) Is there a
seasonal gradient in the di-versity of individual pollen loads, and
is this mediated by plant or pollinator species-level traits?
2 | MATERIALS AND METHODS
2.1 | Study design and data collection
Temperate deciduous forests support many native bee species,
most of which are active as adults only in the spring when the
forest canopy has not closed due to leaf-out, and spring ephemeral
herbs and animal-pollinated trees are in bloom (March–May in our
study system). Our study focused on native bees in the genera
Andrena, Augochlora, Augochloropsis, Bombus, Colletes, Lasioglossum
and Osmia (hereafter “forest bees”), because these genera are known
to include species associated with forest habitat (Harrison et al.,
2018), and we wished to better understand floral resource use by
this understud-ied group. We excluded all male bees and parasitic
species because these bees do not collect pollen (Michener,
2000).
We collected bees at four forested study sites in New Jersey and
Pennsylvania (Table S1). All sites were deciduous forests, located
in the Piedmont ecoregion, with oak–hickory or maple–beech–birch
overstories and a mix of native and non-native plants in the
herba-ceous and shrub layers. Average land cover at a 1 km radius
around each site was 55.9% (±12.5% SE) forest, 16.0% suburban
(±5.0% SE), 14.9% (± 6.7% SE) agricultural, 1.1% (± 0.9% SE) water
and 12.2 (±4.9% SE) wetland matrix habitat.
We visited sites between 23 March 2016 and 20 May 2016. Sites
were visited on separate days by two to three data collectors at a
time. All sampling was conducted between 10:00 and 19:00 on sunny
days when the temperature was high enough for bees to be active
(>14°C), but we otherwise did not standardize sampling between
sites. At each site, we walked over an area of approxi-mately
26,390 m2 (±8,875 m2 SE), both through the understorey
and along forest edges, searching for plants with flowers in
bloom. We sampled bees from as many plant species as we could find,
48 plant species in total (mean = 15.0 ± 5.2 SE plant species per
site; Table S2). We used an insect net to collect any bee we
observed visiting a flower, placing each individual bee into a
separate clean vial. Because our primary interest was in the
polylectic bee spe-cies whose foraging at the individual level has
greater potential to vary, we minimized collecting a highly
abundant bee species, Andrena erigeniae Robertson 1891, which is
narrowly oligolectic on a single plant species, Claytonia virginica
L. We froze all bees in coolers in the field and stored them at
−30°C in a laboratory freezer until processing. Bees were
identified using published revi-sions (see Gibbs, Ascher,
Rightmyer, & Isaacs, 2017 for references; LaBerge, 1986).
We checked whether floral diversity increased during the course
of spring at our study sites by examining how the diversity of
plant species we collected bees from changed over time. We found a
pos-itive relationship, as expected based on other studies in our
region (Figure S1).
2.2 | Pollen analysis
In the laboratory, we removed and stained pollen from the
pollen-collecting structures (scopae or corbiculae) of all female
non-para-sitic forest bees. Because a female bee removes her pollen
load after each foraging trip, pollen loads represent a history of
the plant spe-cies a bee has visited up to the point it was
collected during the cur-rent foraging bout. Additionally, because
female bees actively pack pollen into scopae and corbiculae, the
pollen load also represents the pollen actually chosen by the
foraging female for her offspring's consumption.
We removed pollen from each bee's scopal structures used two
1-mm3 cubes of fuchsin (Kearns & Inouye, 1993). We dabbed each
cube over the scopal hairs or corbiculae of a specimen, including
both legs (for bees with scopae or corbiculae on their legs),
covering as much surface area as possible. We stopped either when
the cube was covered in pollen or there was no pollen left on the
specimen. Fuchsin cubes were then melted onto glass slides using a
hot plate.
We identified pollen to the finest taxonomic level possible
under a light microscope. Pollen identifications were conducted
using a pollen reference library we curated from our study sites.
The library contained a total of 127 plant species, including both
forest and for-est-edge plant species. Pollen reference libraries
were compiled by collecting freshly dehiscent anthers from plants
in bloom at a site and staining pollen from these anthers using
fuchsin gel. We were not able to distinguish some closely related
plant species by their pollen and thus in some cases use genus-,
family- or higher-level groupings (Table S3). Lastly, not all
pollen in bee pollen loads was in our pollen reference libraries.
We used morphospecies classifica-tions for these pollens, but these
morphospecies-level pollens con-stituted only 6.5% of all pollen in
pollen loads. Hereafter, we refer to all pollen groups (of family,
genera, species and morphospecies) as “morphotypes”.
-
| 1161Journal of Animal EcologySMITH eT al.
For each bee, we systematically surveyed about half of all
pol-len grains that had adhered to the surface of the two fuchsin
cubes (thus, we use a standard amount of sampling effort for each
bee). We vertically scanned half of each fuchsin stain at 200x
magnification and recorded all pollen morphotypes we observed in
each scan. We counted the number of scans in which each morphotype
occurred to quantify each pollen morphotype's relative incidence in
the pollen load. To avoid analysing pollen that might be
contamination, we only recorded pollen morphotypes that had at
least five pollen grains in a slide. We excluded individual bees
from analyses if they carried no detectable pollen (n = 25).
2.3 | Classifying species’ traits
We classified bees and plants by the following species-level
traits: bee sociality, bee species-level dietary specialization
(hereafter, lecty), bee body size and plant growth form (Table S4).
To categorize bees by their sociality and lecty, we used a
previously compiled traits database of bees in the northeastern
United States (Bartomeus et al., 2013). We categorized bees as
either solitary or social, and as either oligolectic (pollen
specialists) or polylectic (pollen generalists). We consider
oligolectic species to be those that only collect pol-len from a
single plant family, genus or species. We used the USDA Plants
database (USDA & NRCS, 2018) to classify each plant species
that we collected a bee from by its growth form as either
herbaceous or woody.
We measured bee body size for the majority of species using a
published independent dataset (Cariveau et al., 2016), though for
seven species not in the independent dataset and for all species
from the genus Bombus, we measured bee body size using specimens
from our own data, using the same methods as Cariveau et al.
(2016). Queens and workers in the genus Bombus differed markedly in
body size, so for this genus we measured the body sizes of queens
and workers separately. For between 1 and 10 specimens of each bee
species, we measured intertegular distance, the distance between
the bee's wing bases. Intertegular distance is strongly correlated
with dry body mass, and we used a published equation to convert
intertegular distance to body mass (Cane, 1987).
2.4 | Calculating pollen load diversity
We measured the diversity of pollen in individual pollen loads
using richness.
We also compared different choices of diversity metrics by
re-running our analyses using Hill-Shannon diversity (which
provides a balanced measure of rare and common species) and
Hill-Simpson diversity (which emphasizes common species) in place
of richness (Jost, 2006, Appendix S1: Figure S2).
2.5 | Analysis methods
We examined the distribution of pollen load diversities across
all individual bees in our study (question i) by plotting
histograms of
pollen load richness. We also plotted separate histograms for
each of the nine most common bee species (N ≥ 20) in our data, to
exam-ine how these distributions vary between bee species (Appendix
S1: Figure S3).
We used linear mixed-effects models to examine the importance of
traits and seasonality to foraging bout specialization, modelling
individual pollen load diversity as a function of species’ traits,
day-of-year, and the interactions between traits and day-of-year.
We started with a global model in which the response variable was
the richness of pollen carried by an individual bee. The
explanatory vari-ables included four species-level traits:
sociality (social or solitary), lecty (oligolectic or polylectic),
bee body size, and the growth form of the plant the individual was
foraging from when collected (woody or herbaceous), as well as the
day-of-year of collection and all two-way interactions between
species’ traits and day-of-year. Though three-way interactions
might be important for explaining pollen load diversity, we did not
include these in our model to avoid model over-fitting.
We included site and its interaction with day-of-year as
addi-tional explanatory variables to account for non-independence
of individuals from the same site. To account for non-independence
between individuals of the same species, we included species as a
random intercept and slope effect in the model, allowing the
rela-tionship between day-of-year and pollen load diversity to vary
by species (Gelman & Hill, 2007).
To make model intercepts and interactions interpretable, we
standardized and centred continuous explanatory variables (i.e.
day-of-year and body size). To account for heteroscedasticity, we
loge-transformed pollen load diversity and log10-transformed bee
body size. We used the r package “lme4” to fit mixed models to data
(Bates, Mächler, Bolker, & Walker, 2015; R Core Team,
2018).
We used an exploratory method, AICc, to determine the optimal
fixed structure of model. We compared the AICc value of the global
model with all possible submodels containing the variable
day-of-year (using the “dredge” function in the r package “MuMIn”;
Bartón, 2017). We picked the model with the lowest AICc value.
We checked assumptions of normality and homogeneity of vari-ance
for the global and best-fitting models using diagnostic plots of
model residuals. We checked that explanatory variables did not
covary by calculating the variance inflation factor for each fixed
effect in the final model. None exceeded three, suggesting there is
minimal colinearity between explanatory variables (Zuur, Ieno,
Walker, Saveliev, & Smith, 2009). We estimated the best-fitting
model's goodness-of-fit by calculating the marginal R2, the
variance explained by the fixed effects alone, and the conditional
R2, the vari-ance explained by both the fixed and random effects
(Bartón, 2017; Johnson, 2014; Nakagawa & Schielzeth, 2013).
To assess the significance of variables in the final model, we
used parametric bootstrapping to obtain 95% confidence intervals
around model parameter estimates. Using the R command “boot-Mer,”
we generated 1,000 bootstrapped samples from the model and refit
the model to these samples to obtain bootstrapped pa-rameter
estimates (Bates et al., 2015). We then calculated the
-
1162 | Journal of Animal Ecology SMITH eT al.
2.5 and 97.5 percentile values of the bootstrapped parameter
estimates.
We concluded that a trait affects individual pollen load
diversity (question ii) if the trait was included in the
best-fitting model, and the bootstrapped confidence intervals
around its parameter estimate in that model did not overlap with
zero. Because we centred the variable day-of-year around its mean,
the main effects of all traits that interact with day-of-year are
interpretable as the effect of the trait on pollen load diversity
on the mean day of sampling, April 25 (Schielzeth, 2010).
We concluded that a trait mediated the effect of day-of-year on
pollen load diversity (question iii) if the interaction between a
trait and day-of-year was included in the best-fitting model and if
its bootstrapped confidence intervals around the parameter estimate
in that model did not overlap with zero. If there was a significant
interaction between day-of-year and a categorical trait, then we
ex-amined the confidence intervals around the day-of-year
coefficient for each trait value.
To hedge against the possibility that our results could be due
to an arbitrary choice among equally credible models, we also ran
the parametric bootstrapping analysis on all models within three
AICc values of the best-fitting model (Figure S4).
3 | RESULTS
In total, we removed pollen from 441 individuals of 56 forest
bee species (Table S5).
1. What is the distribution of pollen load diversities across
indi-vidual bees collected while foraging?
Across the entire season, 33.4% of individual bees carried pure
(i.e. single morphotype) pollen loads (Figure 1). The average
richness of a pollen load was 2.28 (±0.07 SE; median = 2). The
distribution of pollen load diversities was right-skewed; most
individuals carried pollen loads with one or a few pollen
morphotypes (Figure 1). Each of the common bee species (N ≥ 20) we
examined also had right‐skewed pollen load
diversity distributions (Figure S3). However, the percentage of
individ-uals carrying pure pollen loads varied between bee species,
as did the average richness of pollen individuals carried (Appendix
S1: Table S6).
2. Do species-level traits affect the diversity of pollen
carried by individuals?
The best-fitting model of individual pollen load diversity
included the main effects of day-of-year, sociality, lecty, body
size, plant growth form and site, and the interactions between
day-of-year and lecty, between day-of-year and plant growth form,
and between day-of-year and body size (Table S7). It explained 38%
of variation in individual pollen load diversity (conditional R2 =
0.383), with 24% of the variation explained by the fixed effects
alone (marginal R
2 = 0.243). There were three other models within three AICc
values of the best-fitting model (Table S7; Figure S4).
Individuals of solitary bee species carried significantly
higher-diversity pollen loads than individuals of social species
(Table 1; Figure 2). Being solitary instead of social was
associated with a 28.7% increase in pollen load diversity (when
reporting effect sizes, all other variables are held at their
reference levels or means). Social bees in the genus Bombus carry
very large pollen loads, which might be relatively undersampled. To
approximately control for differ-ences between social and solitary
bees in pollen load size, we com-pared the pollen load diversities
of social and solitary bees within the genus Lasioglossum (N = 90)
and found that, directionally, the effect of sociality was the same
(Figure S5).
F I G U R E 1 Distribution of individual pollen load
diversities, with diversity measured using richness
Pro
port
ion
of in
divi
dual
s
0 2 4 6 8 10
0.0
0.1
0.2
0.3
0.4
0.5
Pollen load diversity (Richness)
TA B L E 1 Parameter estimates of the best-fitting model of
pollen load richness (loge transformed) and bootstrapped 95%
confidence intervals. Significant parameters are highlighted in
bold. The model intercept represents the model prediction when
categorical variables are held at the reference level and
continuous variables are held at their means (reference level =
solitary, polylectic bees visiting herbaceous plants at site1).
Parameter estimates of categorical variables represent the
difference in the model prediction between the non-reference and
reference levels. Parameter estimates of continuous variables are
interpretable as the change in log(richness) with each standard
deviation of the variable. The variable size is
log10-transformed
Parameter Estimate Lower C.I. Upper C.I.
Intercept 0.765 0.623 0.907
Day‐of‐year 0.121 0.034 0.201
Site2 0.207 0.058 0.368
Site3 0.033 −0.184 0.272
Site4 −0.041 −0.189 0.104
Growth −0.060 −0.175 0.056
Lecty −0.157 −0.446 0.123
Size 0.169 0.096 0.247
Sociality −0.252 −0.419 −0.066
Day‐of‐year:Growth 0.150 0.026 0.276
Day‐of‐year:Lecty −0.336 −0.572 −0.096
Day-of-year:Size 0.059 −0.004 0.128
-
| 1163Journal of Animal EcologySMITH eT al.
Bee body size was significantly positively associated with
pollen load diversity (Figure 2; Table 1), with the largest bees
carrying an estimated 130.6% additional morphotypes of pollen in
their pollen loads compared to the smallest bees.
There was no significant difference in the diversity of pollen
col-lected by individuals belonging to oligolectic and polylectic
species (Table 1). The non-significant difference was driven by
seven individ-uals of one oligolectic bee species, Andrena
erythronii (results when these seven individuals were removed from
the analysis: lecty main effect [C.I.] = −0.35 [−0.67, −0.03]).
Individuals from this species car-ried surprisingly diverse pollen
loads and only a small percentage of host pollen (Figure 3b; Table
S8). The determinations for this bee species were verified using a
published key and reference specimens (LaBerge, 1987).
Plant growth form was included in the best-fitting model, but
was not significantly associated with pollen load diversity (Table
1).
The variable site was also included in the best-fitting model as
a main effect. Individuals from different sites varied in the
diversity of pollen they carried, from an average of 2.1 to 2.6
pollen morpho-types (Table 1).
These results were robust to the use of different diversity
met-rics (Appendix S1; Figure S2) and to the choice of model
(Figure S4).
3. Is there a seasonal gradient in the diversity of individual
pollen loads, and is this mediated by plant or pollinator
species-level traits?
The final model included significant interactions between
day-of-year and lecty and between day-of-year and plant growth form
(Table 1). An interaction between day-of-year and body size was
included in the final model, but it was not significant. The
inter-action between day-of-year and sociality was not included in
the final model (Table 1).
Day-of-year had a greater effect on the diversity of pollen
carried by polylectic bees than on the diversity of pollen carried
by oligolec-tic bees (Table 1; Figure 3). For polylectic bees, the
estimated num-ber of pollen morphotypes a bee carried increased by
0.9 from the first day-of-sampling to the last, an increase of
54.2%; for oligolectic
F I G U R E 2 The effects of body size and sociality on the
richness of pollen collected by individual bees. The solid lines
show the model predictions, back-transformed from the log scale.
For the model prediction, all other variables are held at their
reference levels or means
2.5
5.0
7.5
0.0 0.5 1.0 1.5 2.0
Body mass (mg) (logged)
Indi
vidu
al p
olle
n lo
ad d
iver
sity
SocialitySolitarySocial
F I G U R E 3 The effect of day-of-year on the richness of
pollen collected by individuals of (a) polylectic and (b)
oligolectic species. Individuals in the species Andrena erythronii,
which carried unusually diverse pollen loads for an oligolectic
bee, are highlighted in pink. Lines show model predictions,
back-transformed from the log scale. The dashed line is the model
prediction for bees visiting herbaceous plants, and the solid line
is the model prediction for bees visiting woody plants. For model
predictions, all other variables are held at their reference levels
or means
Polylectic Oligolectic
Apr 01 Apr 15 May 01 May 15 Apr 01 Apr 15 May 01 May 15
2.5
5.0
7.5
Day of year
Indi
vidu
al p
olle
n lo
ad d
iver
sity
GrowthHerbaceous plantWoody plant
BeePolylecticOligolecticAndrena erythronii
(a) (b)
-
1164 | Journal of Animal Ecology SMITH eT al.
bees, the estimated number of pollen morphotypes decreased by
1.5, or 53.6%.
The final model also included a significant interaction between
day-of-year and plant growth form (Table 1). Day-of-year had a
greater effect on the pollen load diversities of bees visiting
trees and shrubs than on the pollen load diversities of bees
visiting her-baceous plants (Table 1; Figure 3). For example, bees
visiting woody plants carried an estimated 1.9 additional pollen
morphotypes from the first day of sampling to the last, an increase
of 163.8%, whereas the diversity of pollen carried by bees visiting
herbaceous plants in-creased by only 54.2%.
We also examined the bootstrapped confidence intervals around
the day-of-year coefficient at each trait level and found that the
slope of the relationship between day-of-year and pollen load
di-versity was significantly positive for polylectic bees visiting
herba-ceous plants (slope [C.I.] = 0.12 [0.03, 0.20]) and woody
plants (slope [C.I.] = 0.27 [0.17, 0.37]). It was not significantly
different from zero for oligolectic bees visiting herbaceous plants
(slope [C.I. = −0.21, [−0.45, 0.02]) or woody plants (slope [C.I.]
= −0.06 [−0.31, 0.18]).
We got qualitatively similar results when analysing other
diversity metrics, and when analysing other top-ranked models
(ΔAICc < 3), with one exception (Figures S2 and S4). The fourth
ranked model did not include an interaction between day-of-year and
plant growth form, which was significant in the main analysis
(Figure S4). It instead included a significant interaction between
day-of-year and body size, with day-of-year having a greater effect
on the diversity of pollen car-ried by large bees than on the
diversity of pollen carried by small bees (Figure S4).
4 | DISCUSSION
Pollinator foraging bout specialization affects patterns of
pollen transfer among plants and thus could affect plant
reproduction. However, most of what we know about it comes from
studies of bees in only two genera, Apis and Bombus. Here, we show
that, across 56 bee species in seven genera, most individuals carry
low-diversity pollen loads, but the percentage of individuals
carrying completely pure (i.e. single morphotype) pollen loads is
much lower than has been found in studies of Apis and Bombus. In
our study, only 34% of all individuals carried pure pollen loads,
compared to >90% in studies of Apis and between 49% and 75% in
studies of Bombus (Free, 1963; Grant, 1950; Leonhardt &
Blüthgen, 2012). Furthermore, we show that foraging bout
specialization is affected by plant and pollinator life-history
traits, as well as seasonality.
In our data, social bee species, which have been the focus of
most previous studies, carry less diverse pollen loads than
solitary bee species, which represent the great majority of bee
species on Earth (Figure 2; Table 1). A potential explanation is
that social bees have a greater capacity for learning than solitary
ones and thus learn more quickly which plant species in their
environment is a profitable one to specialize on (Jones &
Agrawal, 2017). There is evidence in other animal taxa that
sociality and learning capacity are correlated
(Byrne & Bates, 2007), and in bees, this explanation is
supported by two previous studies, both of which found that a
social bee spe-cies was faster to learn to associate a plant
species’ reward with its morphology and as a result was more likely
to exhibit foraging bout specialization (Amaya-Márgez & Wells,
2008; Dukas & Real, 1991). In our study, we also show that
social bees are more likely to exhibit foraging bout specialization
than solitary ones, but unlike previous research, we show that this
pattern exists across individuals from many social and solitary bee
species, 21 social bee species in three genera and 35 solitary bee
species in five genera.
Bee body size also affects foraging bout specialization, with
small bees being more specialized foragers than large bees. Across
the tree of life, body size is one of the major explanatory
variables in ecology (Brown, Gillooly, Allen, Savage, & West,
2004). It affects most aspects of an individual's life history and
ecology, including metabolism, home range and resource requirements
(Brown et al., 2004; Reiss, 1988), all of which are likely to
affect foraging bout spe-cialization. A bee that collects more
pollen and flies farther distances while foraging (Greenleaf,
Williams, Winfree, & Kremen, 2007; Ramalho et al., 1994) will
probably encounter more plant species, just by chance. Thus, larger
bees would be expected to visit more plant species during a
foraging bout, even without presuming they seek out greater pollen
diversity. Studies of monospecific crop fields have found that
large bees are more effective pollinators than small bees because
they deposit more pollen grains onto plant stigmas (Bartomeus,
Cariveau, Harrison, & Winfree, 2018; Larsen, Williams, &
Kremen, 2005). However, our result suggests that, in diverse wild
plant communities, large bees could also transfer a greater
propor-tion of heterospecific pollen between flowers, reducing some
of their effectiveness as pollinators.
Foraging bout specialization in polylectic bee species becomes
less frequent as spring progresses, most likely due to the wider
va-riety of plants blooming later in the season (Anderson &
Hubricht, 1940; Heinrich, 1976; Kochmer & Handel, 1986). Floral
diversity sets an upper limit on pollen load diversity, so that, as
long as generalized foraging is energetically profitable for at
least some individuals, it probably tends to increase the average
diversity of individual pol-len loads within a community. Many
regions with shortened grow-ing seasons exhibit seasonal patterns
in floral diversity similar to the one in our study region, with
floral diversity increasing to one or two peaks as abiotic
conditions become favourable (e.g. Anderson & Hubricht, 1940,
Heinrich, 1976, Kochmer & Handel, 1986, Morales et al., 2005,
Makrodimos et al., 2008, Craine et al., 2012). As a result,
seasonal declines in pollinator foraging bout specialization among
polylectic bees, which in many communities comprise the majority of
bee species (Minckley & Roulston, 2006), are likely to be
common.
In contrast to polylectic bees, oligolectic bees did not become
more generalized in their foraging behaviour over time. With the
ex-ception of individuals from one species, they carried mostly or
only pollen from their host plants, even in late spring when floral
diversity was high (Figure 3b). Surprisingly, oligolectic bees were
not different from polylectic bees in the average diversity of
pollen they carried (Table 1), but this was primarily due to seven
individuals from one
-
| 1165Journal of Animal EcologySMITH eT al.
ostensibly oligolectic species, Andrena erythronii, which
carried un-usually diverse pollen loads. This species has
previously been found to collect non-host pollen when its host
plant is not in bloom, so it is likely this species is more
flexible in its pollen use than suggested by a typical definition
of oligolecty (Michener & Rettenmeyer, 1956). It is also
possible we would have observed a stronger effect of lecty had we
collected a common oligolectic bee in our system, A. erige-niae, in
proportion to its abundance.
Plant growth form also affected seasonal changes in pollinator
foraging bout specialization, with visitors to woody plants being
more strongly affected by day-of-year than visitors to herbaceous
plants. We expected the opposite, that the large floral displays
pro-duced by trees and shrubs would cause pollinators to
specialize, even in late spring when floral diversity is high.
Though polylectic bees were more specialized foragers on woody
plants than on herbaceous ones in early spring, as the spring
progressed bee foraging behaviour on woody and herbaceous plants
became similar (Figure 3a). A possi-ble explanation is that in
early spring, bees are more likely to special-ize on trees, because
large floral displays are particularly attractive when flowers in
the landscape are otherwise sparse. In late spring, when flowers
are more abundant, herbs and trees may be more simi-larly
attractive, and foragers may be less likely to specialize on
trees.
Our result has implications for plant fitness and timing of
repro-duction because plants blooming later in spring were more
likely to be visited by bees that also visited other plant species
during the same foraging bout. This could indicate a potential
reproductive cost to delaying bloom, via reduced pollination
quality (Morales & Traveset, 2008). Other studies have shown
that the presence of co-flowering plant species can reduce plant
fitness by decreasing pollinator for-aging bout specialization
(Campbell & Motten, 1985; Flanagan et al., 2011; Flanagan,
Mitchell, Knutowski, & Karron, 2009), and studies show that
plants in more diverse regions experience more pollen limitation
than plants in less diverse regions (Alonso, Vamosi, Knight,
Steets, & Ashman, 2010; Vamosi et al., 2006). Our result
emphasizes that temporal differences in floral diversity could also
be important. Furthermore, temporal changes in floral diversity
have the poten-tial to act as agents of selection on flowering
time, because plants can avoid periods of high floral diversity by
blooming earlier or later. Future research should investigate to
what extent seasonal changes in pollinator foraging bout
specialization affect plant fitness and tim-ing of
reproduction.
Our methods of estimating foraging bout specialization imposed
two limitations on our study. First, we were not able to identify
all pollen to species, and it is possible, for example, that some
mono-typic pollen loads actually contained multiple pollen species
within the same family or genus. Low taxonomic resolution is a
well-known limitation of pollen load data, and most pollen studies
combine at least some pollen species by genus or family (e.g.
Lopezaraiza-Mikel, Hayes, Whalley, & Memmott, 2007, Tur,
Vigalondo, Trøjelsgaard, Olesen, & Traveset, 2014, Keller et
al., 2015, Wood et al., 2018). In our data, 67% of all pollen in
pollen loads was only resolved to the mor-pho-group (not the
species) level (Table S3), though our pollen mor-pho-groups
contained few species (mean = 3.8 ± 0.58 SE, median = 3).
Second, the bees we collected were at all stages of their
foraging bouts, and bees at the start of their bout would have
visited a smaller number of plants. Though it is possible to obtain
pollen loads from bees that have completed entire foraging trips,
by removing pollen as bees return to their nests, it is not
feasible to do this for a wide array of bee species. These sampling
limitations, despite potentially biasing our absolute estimates of
foraging bout specialization, apply similarly across bees and
therefore should not strongly bias our conclusions regarding the
effects of species’ traits and community-level patterns of
flowering phenology on foraging bout specialization.
In conclusion, we show that specialization at the scale of a
foraging bout is common across a wide diversity of bee species, but
not all bee species are equally likely to exhibit foraging bout
specialization at all times. Bee species that are solitary, large
or polylectic are less likely to exhibit foraging bout
specialization than species that are social, small or oligolectic,
either throughout an entire growing season, or when floral
diversity peaks. A next step for future research is to investigate
how changes in pollinator foraging bout specialization over time
and among species affect plant fitness.
ACKNOWLEDGEMENTS
We thank two anonymous reviewers for providing detailed comments
that improved the manuscript. We also thank mem-bers of the Winfree
laboratory for helpful comments; Cameron Kanterman and Michael
Botros for field and laboratory assistance; and the Xerces Society
for Invertebrate Conservation, the Natural Resources Conservation
Service, and the Garden Club of America for providing funding.
AUTHORS’ CONTRIBUTIONS
C.S. and R.W. conceived the idea for the study. C.S. and L.W.
col-lected data. J.G. identified bee specimens. All authors
contributed to the writing of the manuscript.
DATA ACCESSIBILITY
Data are available from the Dryad Digital Repository: https
://doi.org/10.5061/dryad.2bm34c2 (Smith, Weinman, Gibbs, &
Winfree, 2019).
ORCID
Colleen Smith https://orcid.org/0000-0003-0702-6927
REFERENCES
Alonso, C., Vamosi, J. C., Knight, T. M., Steets, J. A., &
Ashman, T. L. (2010). Is reproduction of endemic plant species
particularly pollen limited in biodiversity hotspots? Oikos,
119(7), 1192–1200. https
://doi.org/10.1111/j.1600-0706.2009.18026.x
-
1166 | Journal of Animal Ecology SMITH eT al.
Amaya-Márgez, M., & Wells, H. (2008). Social complexity and
learning foraging tasks in bees. Caldasia, 30(2), 469–477.
Anderson, E., & Hubricht, L. (1940). A method for describing
and com-paring blooming-seasons. Bulletin of the Torrey Botanical
Club, 67(8), 639–648. https ://doi.org/10.2307/2481328
Barrett, S. C. H., & Harder, L. D. (1996). Ecology and
evolution of plant mating. Trends in Ecology & Evolution,
11(2), 73–79. https ://doi.org/10.1016/0169-5347(96)81046-9
Bartomeus, I., Ascher, J. S., Gibbs, J., Danforth, B. N.,
Wagner, D. L., Hedtke, S. M., & Winfree, R. (2013). Historical
changes in northeast-ern US bee pollinators related to shared
ecological traits. Proceedings of the National Academy of Sciences,
110(12), 4656–4660. https ://doi.org/10.1073/pnas.12185 03110
Bartomeus, I., Cariveau, D. P., Harrison, T., & Winfree, R.
(2018). On the inconsistency of pollinator species traits for
predicting either re-sponse to land-use change or functional
contribution. Oikos, 127(2), 306–315. https
://doi.org/10.1111/oik.04507
Bartón, K. (2017). MuMIN: multi-model inference. R package
version 1.40.0. Retrieved from https ://cran.r-proje ct.org/packa
ge=MuMIn
Bascompte, J., Jordano, P., & Olesen, J. M. (2006).
Asymmetric coevo-lutionary networks facilitate biodiversity
maintenance. Science, 312, 431–433. https ://doi.org/10.1126/scien
ce.1123412
Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015).
Fitting linear mixed-effects models using lme4. Journal of
Statistical Software, 67(1), 1–48. https ://doi.org/10.18637/
jss.v067.i01
Brosi, B. J. (2016). Pollinator specialization: From the
individual to the community. New Phytologist, 210(4), 1190–1194.
https ://doi.org/10.1111/nph.13951
Brown, J. H., Gillooly, J. F., Allen, A. P., Savage, V. M.,
& West, G. B. (2004). Toward a metabolic theory of ecology.
Ecology, 85(7), 1771–1789. https ://doi.org/10.1890/03-9000
Byrne, R. W., & Bates, L. A. (2007). Sociality, evolution
and cogni-tion. Current Biology, 17(16), 714–723. https
://doi.org/10.1016/j.cub.2007.05.069
Campbell, D. R., & Motten, A. F. (1985). The mechanism of
competition for pollination between two forest herbs. Ecology,
66(2), 554–563. https ://doi.org/10.2307/1940404
Cane, J. (1987). Estimation of bee size using intertegular span
(Apoidea). Journal of the Kansas Entomological Society, 60(1),
145–147. https ://doi.org/10.1098/rstb.2011.0050
Cariveau, D. P., Nayak, G. K., Bartomeus, I., Zientek, J.,
Ascher, J. S., Gibbs, J., & Winfree, R. (2016). The allometry
of bee proboscis length and its uses in ecology. PLoS ONE, 11(3),
1–13. https ://doi.org/10.1371/journ al.pone.0151482
Castilla, A. R., Pope, N. S., O’Connell, M., Rodriguez, M. F.,
Treviño, L., Santos, A., & Jha, S. (2017). Adding landscape
genetics and individual traits to the ecosystem function paradigm
reveals the importance of species functional breadth. Proceedings
of the National Academy of Sciences, 114(48), 12761–12766. https
://doi.org/10.1073/pnas.16192 71114
Chittka, L., Thomson, J. D., & Waser, N. (1999). Flower
constancy, insect psychology, and plant evolution.
Naturwissenschaften, 86, 361–377. https ://doi.org/10.1007/s0011
40050636
Craine, J. M., Wolkovich, E. M., Towne, E. G., & Kembel, S.
W. (2012). Flowering phenology as a functional trait in a tallgrass
prairie. New Phytologist, 193, 673–682. https
://doi.org/10.1111/j.1469-8137.2011.03953.x
Cripps, C., & Rust, R. W. (1989). Pollen foraging in a
community of Osmia bees (Hymenoptera: Megachilidae). Environmental
Entomology, 18(4), 582–589. https
://doi.org/10.1093/ee/18.4.582
Danforth, B. N., Cardinal, S., Praz, C., Almeida, E. A. B.,
& Michez, D. (2013). The impact of molecular data on our
understanding of bee phylogeny and evolution. Annual Review of
Entomology, 58(1), 57–78. https ://doi.org/10.1146/annur
ev-ento-120811-153633
Dukas, R., & Real, L. A. (1991). Learning foraging tasks by
bees: A com-parison between social and solitary species. Animal
Behaviour, 42, 269–276. https
://doi.org/10.1016/S0003-3472(05)80558-5
Eckhardt, M., Haider, M., Dorn, S., & Müller, A. (2014).
Pollen mixing in pollen generalist solitary bees: A possible
strategy to complement or mitigate unfavourable pollen properties?
Journal of Animal Ecology, 83(3), 588–597. https
://doi.org/10.1111/1365-2656.12168
Flanagan, R. J., Mitchell, R. J., & Karron, J. D. (2011).
Effects of multi-ple competitors for pollination on bumblebee
foraging patterns and Mimulus ringens reproductive success. Oikos,
120(2), 200–207. https
://doi.org/10.1111/j.1600-0706.2010.18777.x
Flanagan, R. J., Mitchell, R. J., Knutowski, D., & Karron,
J. D. (2009). Interspecific pollinator movements reduce pollen
deposition and seed production in Mimulus ringens (Phrymaceae).
American Journal of Botany, 96(4), 809–815. https
://doi.org/10.3732/ajb.0800317
Free, J. B. (1963). The flower constancy of honeybees. Journal
of Animal Ecology, 32(1), 119–131. https
://doi.org/10.2307/2521
Gegear, R. J., & Laverty, T. M. (2005). Flower constancy in
bumblebees: A test of the trait variability hypothesis. Animal
Behaviour, 69(4), 939–949. https ://doi.org/10.1016/j.anbeh
av.2004.06.029
Gelman, A., & Hill, J. (2007). Data analysis using
regression and multilevel/hierarchical models. Cambridge, UK:
Cambridge University Press.
Gibbs, J., Ascher, J. S., Rightmyer, M. G., & Isaacs, R.
(2017). The bees of Michigan (Hymenoptera: Apoidea: Anthophila),
with notes on distri-bution, taxonomy, pollination, and natural
history. Zootaxa, 4352(1), 1. https ://doi.org/10.11646/ zoota
xa.4352.1.1
Grant, V. (1950). The flower constancy of bees. The Botanical
Review, 16(7), 379–398. https ://doi.org/10.1007/BF028 69992
Greenleaf, S. S., Williams, N. M., Winfree, R., & Kremen, C.
(2007). Bee foraging ranges and their relationship to body size.
Oecologia, 153, 589–596. https
://doi.org/10.1007/s00442-007-0752-9
Harrison, T., Gibbs, J., & Winfree, R. (2018). Forest bees
are replaced in agricultural and urban landscapes by native species
with differ-ent phenologies and life-history traits. Global Change
Biology, 24(1), 287–296. https ://doi.org/10.1111/gcb.13921
Heinrich, B. (1976). Flowering phenologies: Bog, woodland, and
disturbed habitats. Ecology, 57(5), 890–899. https
://doi.org/10.2307/1941055
Hicks, D. M., Ouvrard, P., Baldock, K. C. R., Baude, M.,
Goddard, M. A., Kunin, W. E., … Stone, G. N. (2016). Food for
pollinators: Quantifying the nectar and pollen resources of urban
flower meadows. PLoS ONE, 11(6), 1–37. https
://doi.org/10.1371/journ al.pone.0158117
Johnson, P. C. D. (2014). Extension of Nakagawa &
Schielzeth’s R2GLMM to random slopes models. Methods in Ecology and
Evolution, 5, 944–946. https ://doi.org/10.1111/2041-210X.12225
Jones, P. L., & Agrawal, A. A. (2017). Learning in insect
pollinators and herbivores. Annual Review of Entomology, 62(1),
53–71. https ://doi.org/10.1146/annur ev-ento-031616-034903
Jost, L. (2006). Entropy and diversity. Oikos, 113, 363–375.
https ://doi.org/10.1111/j.2006.0030-1299.14714.x
Kaiser-Bunbury, C. N., Mougal, J., Whittington, A. E., Valentin,
T., Gabriel, R., Olesen, J. M., & Blüthgen, N. (2017).
Ecosystem restoration strengthens pollination network resilience
and function. Nature, 542(7640), 223–227. https
://doi.org/10.1038/natur e21071
Kearns, C. A., & Inouye, D. W. (1993). Techniques for
pollination biologists (1st ed.). Niwot, CO: University Press of
Colorado.
Keller, A., Danner, N., Grimmer, G., Ankenbrand, M., von der
Ohe, K., von der Ohe, W., … Steffan-Dewenter, I. (2015). Evaluating
multiplexed next-generation sequencing as a method in palynology
for mixed pollen samples. Plant Biology, 17, 558–566. https
://doi.org/10.1111/plb.12251
Kochmer, J. P., & Handel, S. N. (1986). Constraints and
competition in the evolution of flowering phenology. Ecology,
56(4), 303–325.
Kunin, W. E. (1993). Sex and the single mustard: Population
density and pollinator behavior effects on seed set. Ecology,
74(7), 2145–2160. https ://doi.org/10.2307/1940859
LaBerge, W. E. (1986). A revision of the bees of the genus
Andrena of the Western Hemisphere. Part XI. Minor subgenera and
subgeneric key. Transactions of the American Entomological Society,
111, 440–567.
-
| 1167Journal of Animal EcologySMITH eT al.
LaBerge, W. E. (1987). A revision of the bees of the genus
Andrena of the Western Hemisphere. Part XII. Subgenera Leucandrena,
Ptilandrena, Scoliandrena, and Melandrena. Transactions of the
American Entomological Society, 112, 191–248.
Larsen, T. H., Williams, N. M., & Kremen, C. (2005).
Extinction order and altered community structure rapidly disrupt
eco-system functioning. Ecology Letters, 8(5), 538–547. https
://doi.org/10.1111/j.1461-0248.2005.00749.x
Leonhardt, S. D., & Blüthgen, N. (2012). The same, but
different: Pollen foraging in honeybee and bumblebee colonies.
Apidologie, 43(4), 449–464. https
://doi.org/10.1007/s13592-011-0112-y
Lopezaraiza-Mikel, M. E., Hayes, R. B., Whalley, M. R., &
Memmott, J. (2007). The impact of an alien plant on a native
plant-pollinator net-work: An experimental approach. Ecology
Letters, 10(7), 539–550. https
://doi.org/10.1111/j.1461-0248.2007.01055.x
Lucas, A., Bodger, O., Brosi, B. J., Ford, C. R., Forman, D. W.,
Greig, C., … de Vere, N. (2018). Generalisation and specialisation
in hoverfly (Syrphidae) grassland pollen transport networks
revealed by DNA metabarcoding. Journal of Animal Ecology, 87(4),
1008–1021. https ://doi.org/10.1111/1365-2656.12828
Makrodimos, N., Blionis, G. J., Krigas, N., & Vokou, D.
(2008). Flower morphology, phenology and visitor patterns in an
alpine commu-nity on Mt Olympos, Greece. Flora, 203, 449–468. https
://doi.org/10.1016/j.flora.2007.07.003
Michener, C. D. (2000). Bees of the world. Baltimore, MD: The
Johns Hopkins University Press.
Michener, C. D., & Rettenmeyer, C. W. (1956). The ethology
of Andrena erythronii with comparative data on other species
(Hymenoptera, Andrenidae). The University of Kansas Science
Bulletin, 37(16), 645–684.
Minckley, R. L., & Roulston, T. H. (2006). Incidental
mutualisms and pol-len specialization among bees. In N. Waser,
& J. Ollerton (Eds.), Plant-pollinator interactions: From
specialization to generalization (pp. 69–98). Chicago, IL:
University of Chicago Press.
Morales, C., & Traveset, A. (2008). Interspecific pollen
transfer: Magnitude, prevalence and consequences for plant fitness.
Critical Reviews in Plant Sciences, 27(4), 221–238. https
://doi.org/10.1080/07352 68080 2205631
Morales, M. A., Dodge, G. J., & Inouye, D. W. (2005). A
phenological mid-domain effect in flowering diversity. Oecologia,
142(1), 83–89. https ://doi.org/10.1007/s00442-004-1694-0
Motten, A. F. (1986). Pollination ecology of the spring
wildflower com-munity of a temperate deciduous forest. Ecological
Monographs, 56(1), 21–42. https ://doi.org/10.2307/2937269
Müller, A. (1996). Host-plant specialization in western
Palearctic Anthidine bees (Hymenoptera: Apoidea: Megachilidae).
Ecological Monographs, 66(2), 235–257. https
://doi.org/10.2307/2963476
Müller, A., Diener, S., Schnyder, S., Stutz, K., Sedivy, C.,
& Dorn, S. (2006). Quantitative pollen requirements of solitary
bees: Implications for bee conservation and the evolution of
bee-flower relationships. Biological Conservation, 130(4), 604–615.
https ://doi.org/10.1016/j.biocon.2006.01.023
Nakagawa, S., & Schielzeth, H. (2013). A general and simple
method for obtaining R2 from generalized linear mixed-effects
mod-els. Methods in Ecology and Evolution, 4, 133–142. https
://doi.org/10.1111/j.2041-210x.2012.00261.x
Ogilvie, J. E., & Forrest, J. R. (2017). Interactions
between bee forag-ing and floral resource phenology shape bee
populations and com-munities. Current Opinion in Insect Science,
21, 75–82. https ://doi.org/10.1016/j.cois.2017.05.015
Pornon, A., Andalo, C., Burrus, M., & Escaravage, N. (2017).
DNA me-tabarcoding data unveils invisible pollination networks.
Scientific Reports, 7(1), 1–11. https
://doi.org/10.1038/s41598-017-16785-5
Praz, C. J., Müller, A., & Dorn, S. (2008). Specialized bees
fail to de-velop on non-host pollen: Do plants chemically protect
their pollen? Ecology, 89(3), 795–804. https
://doi.org/10.1890/07-0751.1
R Core Team. (2018). R: A language and environment for
statistical com-puting. Vienna, Austria: R Foundation for
Statistical Computing. Retrieved from https ://www.r-proje
ct.org/
Ramalho, M., Giannini, T. C., Malagodi-Braga, K. S., &
Imperatriz-Fonseca, V. L. (1994). Pollen harvest by stingless bee
foragers (Hymenoptera, Apidae, Meliponinae). Grana, 33(4–5),
239–244. https ://doi.org/10.1080/00173 13940 9429005
Reiss, M. (1988). Scaling of home range size: Body size,
metabolic needs and ecology. Trends in Ecology and Evolution, 3(3),
85–86. https ://doi.org/10.1016/0169-5347(88)90025-0
Schielzeth, H. (2010). Simple means to improve the
interpretability of regression coefficients. Methods in Ecology and
Evolution, 1(2), 103–113. https
://doi.org/10.1111/j.2041-210X.2010.00012.x
Smith, C., Weinman, L. R., Gibbs, J., & Winfree, R. (2019).
Data from: Specialist foragers in forest bee communities are small,
social or emerge early. Dryad Digital Repository, https
://doi.org/10.5061/dryad.2bm34c2
Tur, C., Vigalondo, B., Trøjelsgaard, K., Olesen, J. M., &
Traveset, A. (2014). Downscaling pollen-transport networks to the
level of in-dividuals. Journal of Animal Ecology, 83(1), 306–317.
https ://doi.org/10.1111/1365-2656.12130
USDA & NRCS. (2018). The PLANTS database. Greensboro, NC:
National Plant Data Team. http://plants.usda.gov
Vamosi, J. C., Knight, T. M., Steets, J. A., Mazer, S. J., Burd,
M., & Ashman, T.-L. (2006). Pollination decays in biodiversity
hotspots. Proceedings of the National Academy of Sciences, 103(4),
956–961. https ://doi.org/10.1073/pnas.05071 65103
Waser, N., Chittka, L., Price, M. V., Williams, N. M., &
Ollerton, J. (1996). Generalization in pollination systems, and why
it matters. Ecology, 77(4), 1043–1060. https
://doi.org/10.2307/2265575
Williams, N. M. (2003). Use of novel pollen species by
specialist and gen-eralist solitary bees (Hymenoptera:
Megachilidae). Oecologia, 134, 228–237. https
://doi.org/10.1007/s00442-002-1104-4
Wood, T. J., Gibbs, J., Rothwell, N., Wilson, J. K., Gut, L.,
Brokaw, J., & Isaacs, R. (2018). Limited phenological and
dietary overlap between bee communities in spring flowering crops
and herbaceous en-hancements. Ecological Applications, 28(7),
1924–1934. https ://doi.org/10.1002/eap.1789
Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A., &
Smith, G. M. (2009). Mixed effects models and extensions in ecology
with R (1st ed.). New York, NY: Springer.
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section at the end of the article.
How to cite this article: Smith C, Weinman
L, Gibbs J, Winfree R. Specialist foragers in forest bee
communities are small, social or emerge early. J Anim Ecol.
2019;88:1158–1167. https ://doi.org/10.1111/1365-2656.13003
-
1
Appendix S1: Supplementary methods, results, tables and figures
1
2
Supplementary methods and results 3
Pollen load diversity measurements using Hill numbers 4
We used a Hill numbers approach to measuring diversity because
this is theoretically 5
preferable to other diversity metrics and provides a way to
explicitly indicate the emphasis given 6
to rare versus common species (Jost, 2006). Hill diversities are
based on one general equation, of 7
which species richness and forms of the traditional Shannon and
Simpson indices are special 8
cases (Jost, 2006). Specifically, Hill-Shannon is the
traditional Shannon index exponentiated, 9
and Hill-Simpson is the inverse of the traditional Simpson
index. All Hill diversities are in the 10
same units, called the effective number of species, which is the
number of species that would be 11
present in a perfectly even community (i.e., one in which each
species was represented by the 12
same number of individuals) with the same diversity metric value
as the one being studied. 13
We calculated pollen load diversity with Hill numbers using
incidence data (that is, the 14
frequencies with which pollen morphotypes occur across multiple
microscope scans of a slide). 15
For the ith pollen morphotype in a pollen load, !! is the
proportion of scans containing that 16
morphotype, and !! = !!!!!! , the relative incidence of the
pollen morphotype in the pollen load. 17
Hill numbers based off incidence data are described by the
following equation (Chao et al., 18
2014): 19
! ! = p!!
!
!!!
!!!!
for q ≥ 0 q ≠ 1
exp − p! lnp!!
!!! for q = 1
-
2
where S = the number of pollen morphotypes in the individual
pollen load and q = the diversity 20
order, which is used to indicate the Hill number’s sensitivity
to rare or common species. When q 21
= 0, ! ! is species richness, which is sensitive to rare
species; when q =1, ! ! is Hill-Shannon 22
diversity, which provides a balanced measure of rare and common
species; and when q=2, ! ! is 23
Hill-Simpson diversity, which is more sensitive to common
species. 24
We use a diversity order of q = 0 (i.e., species richness) for
our main analysis. To ensure 25
our results were robust to our choice of diversity metric, we
also reran our analyses using 26
diversity orders of q = 1 (i.e., Hill-Shannon diversity) and q =
2 (i.e., Hill-Simpson diversity). 27
Our results were qualitatively the same between the three
different diversity orders (Fig. 28
S2). In the analyses of all three diversity orders, there were
significant main effects of site, day-29
of-year, sociality and body size, and significant interactions
between day-of-year and lecty and 30
between day-of-year and plant growth form (Fig. S2). 31
32
Comparing the distributions of individual pollen load
diversities across bee species 33
We examined the distributions of individual pollen load
diversities for nine common bee 34
species in our data (N ≥ 20). All species we examined had right
skewed pollen load diversity 35
distributions, with most individuals carrying low diversity
pollen loads (Fig. S3). However, the 36
percentage of individuals carrying pure pollen loads varied
widely among species, ranging from 37
85.2% of Andrena zizeae to only 4.0% of A. nasonii (Table S6).
The median richness of 38
individual pollen loads ranged from one (in A. zizeae, C.
inaequalis, and L. subvirdatum) to three 39
(in B. bimaculatus; Table S6). 40
41
42
-
3
References 43
Chao, A., Gotelli, N. J., Hsieh, T. C., Sander, E. L., Ma, K.
H., Colwell, R. K., & Ellison, A. M. 44
(2014). Rarefaction and extrapolation with Hill numbers: a
framework for sampling and 45
estimation in species diversity studies. Ecological Monographs,
84(1), 45–67. 46
doi:10.1890/13-0133.1 47
Jost, L. (2006). Entropy and diversity. Oikos, 113, 363–375.
doi:10.1111/j.2006.0030-48
1299.14714.x 49
50
-
4
Supplementary tables 51 52 Table S1. Locations of study sites.
53 54 Site Name Location Latitude Longitude Hutcheson Memorial
Forest USA, NJ, Somerset County 40.5006 -74.5639 Institute Woods
USA, NJ, Mercer County 40.3240 -74.6631 Sourland Mountain Preserve
USA, NJ, Somerset County 40.4731 -74.6956 Bowman's Hill USA, PA,
Bucks County 40.3285 -74.9432
55 56
-
5
Table S2. Plant species and the number of bees collected off of
each. 57
Plant species Bees collected Acer negundo 12 Acer platanoides 1
Acer rubrum 38 Anemone canadensis 3 Aquilegia canadensis 5 Aronia
arbutifolia 2 Barbarea vulgaris 54 Cardamine impatiens 3 Cerastium
nutans 1 Cerastium spp. 1 Cercis canadensis 4 Chrysogonum
virginianum 2 Cornus alternifolia 8 Cornus florida 3 Dicentra
cucullaria 8 Dodecatheon meadia 2 Eleagnus umbellata 6 Erythronium
americanum 5 Euphorbia esula 9 Geranium maculatum 11 Glechoma
hederacea 10 Hepatica nobilis_var._obtusa 1 Hydrophyllum
virginianum 11 Lamium purpureum 2 Lonicera morrowii 4 Malus pumila
7 Malus spp. 19 Mertensia virginica 38 Packera aurea 8 Physocarpus
opulifolius 3 Polemonium reptans 2 Prunus americana 7 Prunus avium
45 Pyrus calleryana 9 Ranunculus bulbosus 11 Ranunculus ficaria 8
Rhododendron calendulaceum 1 Rhododendron periclymenoides 2
Rhododendron prinophyllum 2 Rubus pensilvanicus 2
-
6
Stylophorum diphyllum 6 Taraxacum officinale 29 Tradescantia
virginiana 2 Trifolium repens 2 Viburnum opulus 1 Viola spp. 2
Zizia aurea 30
58
59
-
7
Table S3. Pollen morphotypes (n=46) found on the bees in our
study. Thirty pollen morphotypes 60
were resolved to the species level. The rest are grouped by
clade, family or genus. For each 61
pollen morphotype, we list the taxonomic level of the grouping
and the mean relative incidence 62
of the morphotype across all pollen loads. The five most common
pollen morphotypes were Acer 63
(17.9%), Rosaceae (16.4%), Brassicaceae (12.0%), Mertensia
virginica (10.5%) and Zizea aurea 64
(6.7%). 65
66
Pollen group Taxonomic level of grouping Mean (%) SE (%)
Alliaria petiolata / Euphorbia esula Clade 0.99 0.27 Euonymus alata
/ Viburnum prunifolium Clade 0.59 0.19 Anemone / Aquilegia Family
1.64 0.54 Asteraceae Family 1.60 0.49 Betulaceae Family 0.32 0.24
Brassicaceae Family 12.04 1.25 Duchesnea indica/ Fragaria
virginiana Family 0.46 0.19 Lauraceae Family 0.53 0.25
Ranunculaceae Family 3.02 0.68 Rosaceae Family 16.42 1.38 Acer
Genus 17.92 1.52 Carex Genus 0.07 0.07 Cerastium Genus 0.26 0.22
Lonicera Genus 0.61 0.28 Magnolia Genus 0.07 0.05 Rhododendron
Genus 1.20 0.48 Viola Genus 0.27 0.19 Ajuga reptans Species 0.17
0.12 Cercis canadensis Species 1.49 0.45 Claytonia virginica
Species 0.72 0.29 Cornus alternifolia Species 1.60 0.54 Cornus
florida Species 0.92 0.25 Dicentra cucullaria Species 0.55 0.22
Dodecatheon meadia Species 0.50 0.30 Duchesnea indica Species 0.08
0.09 Eleagnus umbellata Species 0.61 0.25 Erythronium americanum
Species 0.50 0.25 Geranium maculatum Species 0.25 0.22 Glechoma
hederacea Species 0.35 0.17
-
8
Hydrophyllum virginianum Species 1.33 0.44 Lamiaceae Species
0.01 0.01 Lamium purpureum Species 0.09 0.07 Liriodendron
tulipifera Species 1.11 0.31 Mertensia virginica Species 10.49 1.33
Physocarpus opulifolius Species 0.08 0.08 Polemonium reptans
Species 0.40 0.24 Rubus pensilvanicus Species 0.06 0.06 Sassafras
albidum Species 0.04 0.03 Saxifraga pensylvanica Species 0.02 0.02
Saxifraga virginensis Species 0.11 0.11 Stylophorum diphyllum
Species 0.78 0.28 Taraxacum officinale Species 5.41 0.89
Tradescantia virginiana Species 0.48 0.20 Trifolium repens Species
0.30 0.15 Viburnum opulus Species 0.06 0.07 Zizia aurea Species
6.74 1.16 Various morphospecies NA 6.72 0.75
67 68 69
-
9
Table S4. Species’ traits used in the analysis, the number of
individual bees and bee species in 70
each trait category, as well as the number of plant species that
individual bees in each trait 71
category were collected from. Two individuals from two
facultatively social bee species were 72
categorized as social. 73
74 75
Trait Values Individual bees Bee species Plant species
Sociality Social 104 21 33
Solitary 339 35 42
Lecty Oligolectic 47 7 14
Polylectic 396 49 47
Growth form Herbaceous 266 50 28
Woody 177 30 21
Body size Range: 1.2 to 121.5 mg 443 56 49 76
-
10
Table S5. List of bee species we removed pollen from and the
number of individuals of each 77
species collected. 78
79 Species Authority n Andrena alleghaniensis Viereck, 1907 1
Andrena banksi Malloch, 1917 3 Andrena carlini Cockerell, 1901 26
Andrena cornelli Viereck, 1907 4 Andrena cressonii Robertson, 1891
12 Andrena distans Provancher, 1888 1 Andrena erigeniae Robertson,
1891 3 Andrena erythronii Robertson, 1891 7 Andrena fenningeri
Viereck, 1922 6 Andrena forbesii Robertson, 1891 6 Andrena heraclei
Robertson, 1897 2 Andrena hippotes Robertson, 1895 3 Andrena
imitatrix Cresson, 1872 24 Andrena integra Smith, 1853 2 Andrena
mandibularis Robertson, 1892 13 Andrena miserabilis Cresson, 1872
12 Andrena nasonii Robertson, 1895 24 Andrena perplexa Smith, 1853
1 Andrena pruni Robertson, 1891 5 Andrena robertsonii Dalla Torre,
1896 2 Andrena rugosa Cockerell, 1906 14 Andrena vicina Smith, 1853
4 Andrena ziziae Robertson, 1891 27 Augochlora pura (Say, 1837) 16
Augochloropsis metallica (Fabricius, 1793) 4 Bombus bimaculatus
Cresson, 1863 24 Bombus griseocollis (De Geer, 1773) 6 Bombus
impatiens Cresson, 1863 1 Bombus perplexus Cresson, 1863 5 Colletes
inaequalis Say, 1837 24 Lasioglossum abanci (Crawford, 1932) 2
Lasioglossum birkmanni (Crawford, 1906) 2 Lasioglossum cattellae
(Ellis, 1913) 1 Lasioglossum coeruleum (Robertson, 1893) 5
Lasioglossum coriaceum (Smith, 1853) 1 Lasioglossum cressonii
(Robertson, 1890) 9 Lasioglossum foxii (Robertson, 1895) 19
-
11
Lasioglossum fuscipenne (Smith, 1853) 1 Lasioglossum gotham
Gibbs, 2011 2 Lasioglossum illinoense (Robertson, 1892) 1
Lasioglossum imitatum (Smith, 1853) 6 Lasioglossum obscurum
(Robertson, 1892) 1 Lasioglossum pilosum (Smith, 1853) 1
Lasioglossum quebecense (Crawford, 1907) 6 Lasioglossum
subviridatum (Cockerell, 1938) 21 Lasioglossum truncatum
(Robertson, 1901) 6 Lasioglossum versatum (Robertson, 1902) 5
Lasioglossum viridatum (Lovell, 1905) 1 Lasioglossum weemsi
(Mitchell, 1960) 1 Osmia atriventris Cresson, 1864 15 Osmia
bucephala Cresson, 1864 1 Osmia collinsiae Robertson, 1905 2 Osmia
conjuncta Cresson, 1864 1 Osmia georgica Cresson, 1878 3 Osmia
lignaria Say, 1837 1 Osmia pumila Cresson, 1864 47
80 81
-
12
Table S6. The nine most common bee species in our dataset (N ≥
20), with the number of 82
individuals collected, the percentage of individuals carrying
pure (i.e. single morphotype) pollen 83
loads, and the median richness of pollen loads. 84
Bee species N % pure pollen loads Median richness Osmia pumila
47 17.0 2.0 Andrena ziziae 27 85.2 1.0 Andrena carlini 26 7.7 2.5
Andrena imitatrix 24 45.8 2.0 Andrena nasonii 24 4.2 2.0 Bombus
bimaculatus 24 25.0 3.0 Colletes inaequalis 24 58.3 1.0
Lasioglossum subviridatum 21 57.1 1.0
85
-
13
Table S7. Results of model selection: AICc values, ΔAICc values
and weights of models within 86
ten AICc values of the best-fitting model. All models included
day-of-year as a main fixed effect 87
and bee species as a random intercept and slope effect. 88
day site growth lecty size sociality day: site day: growth day:
lecty day: size day: sociality AICc ΔAICc wi + + + + + +
+ + +
606.53 0.00 0.251
+ + + + + +
+ +
607.78 1.25 0.134 + + + + + +
+ + + + 608.69 2.16 0.085
+ +
+ + +
+ +
609.32 2.79 0.062 + + + + + +
+ +
+ 609.93 3.40 0.046
+ +
+ + +
+ + + 610.44 3.90 0.036 + + + + + +
+ +
610.47 3.94 0.035
+ +
+ + +
+
611.22 4.69 0.024 + + +
+ +
+
+
611.43 4.90 0.022
+ + + + + +
+ + + 611.93 5.40 0.017 + + + + + +
+
+
611.94 5.41 0.017
+ + + + + +
+
612.25 5.71 0.014 + +
+ + + +
+ +
612.28 5.75 0.014
+ +
+ +
+ +
612.30 5.77 0.014 + +
+ + +
+
+ 612.40 5.87 0.013
+
+ + +
+ +
612.46 5.93 0.013 + + + + +
+ + +
612.62 6.09 0.012
+ + + + + + + + + +
612.63 6.10 0.012 + + +
+ +
+
+ + 613.05 6.52 0.010
+ + + + + + + + +
613.64 7.11 0.007 + + + + + +
+
+ + 613.70 7.17 0.007
+
+ + + +
+ +
613.71 7.18 0.007 + + + + + +
+
+ 613.72 7.19 0.007
+
+ + +
+
613.86 7.32 0.006 + +
+ + + +
+
613.91 7.38 0.006
+
+ + +
+ + + 613.99 7.45 0.006 + + +
+ +
+
614.13 7.60 0.006
+
+ + + +
+ + +
614.20 7.67 0.005 + + + + +
+ +
614.23 7.70 0.005
+ + + + + + +
+ +
614.38 7.85 0.005 + +
+ + + +
+ + + 614.38 7.85 0.005
+ + + + + + + + + + + 614.81 8.27 0.004 + + + + +
+ +
614.83 8.30 0.004
+
+ + + +
+ +
614.84 8.31 0.004 + + + + + +
+
614.89 8.36 0.004
+ +
+ +
+
615.00 8.47 0.004 +
+ + + +
+
615.01 8.48 0.004
+
+ + +
+
+ 615.31 8.78 0.003 + + +
+
+
+
615.42 8.89 0.003
+
+ + + +
+ + + 615.50 8.96 0.003
-
14
+ +
+ +
+
615.55 9.02 0.003 + + +
+ +
+
+ 615.65 9.12 0.003
+ + + + + + + + +
+ 615.81 9.28 0.002 + + + + + + +
+
615.85 9.31 0.002
+ +
+ + + +
+
+ 615.87 9.33 0.002 + + +
+ +
+
615.94 9.41 0.002
+ +
+ + +
+
615.96 9.43 0.002 + +
+
+
616.19 9.65 0.002
+
+ + + +
+ + + + 616.32 9.79 0.002 + +
+ + +
+
616.38 9.85 0.002
+ + + + + + + + + + 616.42 9.89 0.002 89
90
-
15
Table S8. Oligolectic bee species collected, their host plants,
and the pollen found in their pollen 91
loads. Pollen taxa from a bee species’ host plant is highlighted
in gray. 92
Bee species N Host plant Pollen Mean (%) S.E. Andrena cornelli 4
Rhododendron spp. Rhododendron spp. 100.00 0.00 Andrena distans 1
Geranium maculatum Geranium maculatum 100.00 NA Andrena erigeniae 4
Claytonia virginica Claytonia virginica 75.76 15.30
Ranunculus spp. 24.24 15.30
Andrena erythronii 7 Erythronium spp. Acer spp. 31.19 9.16
Taraxacum officinale 25.78 7.54
Rosaceae 16.68 7.98
Stylophorum diphyllum 10.95 7.11
Erythronium americanum 8.57 7.05
Claytonia virginica 4.63 4.15
Betulaceae 1.05 1.05
Morphospecies1 0.71 0.71
Brassicaceae 0.43 0.43
Andrena integra 2 Cornus (Swida) spp. Cornus alternifolia 95.45
4.55
Zizia aurea 4.55 4.55
Andrena ziziae 27 Zizia spp. Zizia aurea 98.35 0.82
Morphospecies2 1.45 0.81
Rosaceae 0.21 0.21
Osmia georgica 3 Asteraceae Asteraceae 81.79 11.65
Ranunculus 16.76 12.31
Zizia aurea 1.45 1.45 93
94
95
-
16
Supplementary Figures 96
97
98
Figure S1. Increase in floral diversity over time, with floral
diversity measured as the rarefied 99
richness of plant species bees were collected from. To estimate
floral diversity, we used all of 100
our data (N= 1301 bees) including males, parasitic species, and
non-forest bees. We used 101
rarefaction to estimate plant species richness at an equal
sample size (n = 26) for each site-date. 102
We excluded 10 site-dates which did not meet a minimum sample
size of 25 bees. 103
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
Apr 01 Apr 15 May 01 May 15
24
68
Date
Flor
al d
ivers
ity (R
aref
ied
richn
ess)
-
17
104
Figure S2. Results of parametric boostrapping analyses when
Richness, Hill-Shannon diversity 105
and Hill-Simpson diversity (all loge-transformed) are used to
measure pollen load diversity. The 106
points represent parameter estimates and lines represent
bootstrapped 95% confidence intervals. 107
Points are in red if the bootstrapped confidence intervals do
not overlap with zero. 108
109
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
Hill−Shannon Hill−Simpson Richness
−0.5 0.0 0.5 −0.5 0.0 0.5 −0.5 0.0 0.5
Day−of−year
Day−of−year:Growth
Day−of−year:Lecty
Day−of−year:Size
Growth
Intercept
Lecty
Site2
Site3
Site4
Size
Sociality
Estimate
Parameter Significant
●
●
NoYes
-
18
110
Figure S3. The distributions of individual pollen load
diversities of common species (N ≥ 20), 111
with pollen load diversity measured using species richness.
112
113
114 115
Osmia pumila N= 47
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Andrena ziziae N= 27
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Andrena carlini N= 26
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Andrena imitatrix N= 24
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Andrena nasonii N= 24
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Bombus bimaculatus N= 24
0 2 4 6 80.
00.
20.
40.
60.
8Colletes inaequalis
N= 24
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Lasioglossum subviridatum N= 21
0 2 4 6 8
0.0
0.2
0.4
0.6
0.8
Pollen load diversity (Richness)
Freq
uenc
y
-
19
116 Figure S4. Results of parametric boostrapping analyses of
the top four ranked models (all ΔAICc 117
< 3) explaining pollen load richness (loge-transformed). The
points represent parameter estimates 118
and lines represent bootstrapped 95% confidence intervals.
Points are in red if the bootstrapped 119
confidence intervals do not overlap with zero. Results were
qualitatively similar between the top 120
four ranked models, with one exception. The fourth-ranked model
did not include the variable 121
plant growth form, or its interaction with day-of-year; this
interaction was significant in the three 122
other top-ranked models. The fourth-ranked model instead
included a significant interaction 123
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
Rank 3 Rank 4
Rank 1 Rank 2
−0.5 0.0 0.5 −0.5 0.0 0.5
Day−of−yearDay−of−year:Growth
Day−of−year:LectyDay−of−year:Size
Day−of−year:SocialGrowth
InterceptLectySite2Site3Site4Size
Sociality
Day−of−yearDay−of−year:Growth
Day−of−year:LectyDay−of−year:Size
Day−of−year:SocialGrowth
InterceptLectySite2Site3Site4Size
Sociality
Estimate
Parameter Significant
●
●
NoYes
-
20
between day-of-year and body size, with day-of-year having a
greater effect on the diversity of 124
pollen carried by large bees than on the diversity of pollen
carried by small bees. 125
126
127
-
21
128
Figure S5. Box plots illustrating the difference in the richness
of pollen carried by individuals of 129
solitary (N=27) and social bee (N=55) species in the genus
Lasioglossum. To minimize co-130
linearity between body size and sociality, we excluded the eight
smallest individuals of social 131
Lasioglossum, which were outliers in terms of their body size
[log10(body size) < 0.40]. The 132
boxes encompass the second and third quartiles of the data and
the thick black line is the median. 133
The plot whiskers extend to 1.5 times the interquartile range
and circles represent outliers. 134
●
●●
●
●
Solitary Social
12
34
56
7
Pollen load diversities of social and solitary Lasioglossum
Sociality
Polle
n lo
ad ri
chne
ss
