!
!
!
!
!
WILD BEE COMMUNITY RESPONSES TO FARM MANAGEMENT PRACTICES,
WILDFLOWER RESTORATIONS, AND LANDSCAPE COMPOSITION
By
Emily A. May

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Entomology—Master of Science
2015

ABSTRACT
WILD BEE COMMUNITY RESPONSES TO FARM MANAGEMENT PRACTICES,
WILDFLOWER RESTORATIONS, AND LANDSCAPE COMPOSITION
By
Emily A. May
Conservation strategies for wild bee pollinators require information on community and
species-level responses to environmental resources and stressors, including bees’ responses to
habitat composition and agricultural practices at different scales. The objectives of this research
were to 1) determine the direction and variability of wild bee community responses in blueberry
fields to pest management practices and landscape composition by revisiting previously-sampled
farms where pesticide use has increased over recent years; and 2) to determine which habitats in
Michigan farm landscapes are most suitable for soil-nesting bees, and whether bees nest
preferentially in wildflower plantings over nearby unrestored habitats. The abundance, richness,
diversity, and community composition of wild bees exhibited strong negative responses to
insecticide program risk at the field scale over five years of sampling in 15 blueberry fields. In
general, solitary bee species had more negative responses to insecticide risk than social species.
Bees responded most strongly to landscape composition at the smallest scale studied (300 m).
Wild bee abundance declined with decreasing area of forest and the associated increase in settled
areas, while bee richness and diversity declined primarily with insecticide risk. Using emergence
traps, I found that wildflower plantings generally support a greater abundance of soil-nesting
bees than other habitat types in the surrounding landscape. Bee nesting abundance was higher in
mature wildflower restorations than newly-established plantings, indicating that wildflower
restorations are an effective conservation tool for building bee populations over time through the
provision of both floral and nesting resources.

Copyright by
EMILY A. MAY
2015

Dedicated to
Neil R. Taylor, Jr.

iv

ACKNOWLEDGMENTS
Many people have helped and supported me through the design and execution of this
project. Thank you to Rufus Isaacs, for believing in me and this project from the outset. Thanks
for always making time for your students and for maintaining such a positive and fun lab
environment. Thank you also to the members of my committee, Doug Landis and Lars Brudvig,
for your support and guidance, and to Julianna Wilson for your willingness to share data,
protocols, and advice in support of this project.
Many thanks to Karlis Galens, Dennis Hartmann, John Calsbeek, Jim Getzoff, Gary and
Beverlee DeJonge, Joe DeGrandchamp, Kelly Bowerman, Dave Calvano, Craig Tiles, Jan Earl
Woods, Larry Bodtke, Curt and Vicki Carini, RJ Rant, Judy Rant, Doug Wassink, Dave Stansby,
Eric Kamphuis, Scott Kamphuis, Joyce Higgins, Paul Rood, George Fritz, Bill Fritz, Joe Leduc,
Trevor Nichols Research Center, Kalamazoo Nature Center, and Southwest Michigan Land
Conservancy for allowing me access to field sites. This research would not have been possible
without your cooperation and support.
Thanks to my research assistants Ashley McNamara, Clara Stuligross, Nicholas Blodgett,
and Kari Grebe, for your long hours in the field and patience with emergence trapping. Thanks to
Keith Mason for all of your help and your endless optimism. Many thanks to Jason Gibbs for
identifying hundreds of bees with no complaints and for providing valuable insights on bee
biology, taxonomy, and life history traits. I am also indebted to Hillary Sardiñas for her
generosity in sharing emergence trapping protocols and insights. Thanks to Sieg Snapp for the
use of her lab and equipment to run the soil texture analyses.

v

This project received funding from a Michigan State University Plant Sciences
Fellowship, NCR-SARE, North Central Regional IPM Center, Coalition for Urban/Rural
Environmental Stewardship, and Syngenta.
Thanks to my parents for your constant support, advice, and enthusiasm for bees,
gardening, and soil. Most of all, thanks to my husband Daniel and our friendly mutt Arthur for
keeping me happy and healthy through this process.

vi

TABLE OF CONTENTS

LIST OF TABLES................................................................................................................. ix
LIST OF FIGURES............................................................................................................. xiii
KEY TO ABBREVIATIONS............................................................................................. xvii
CHAPTER 1: INTRODUCTION.......................................................................................... 1
Biodiversity-ecosystem functioning.......................................................................... 2
Importance of pollination in managed and natural systems...................................... 4
Overview of pollination declines............................................................................... 5
Bee communities and ecological diversity................................................................ 7
Bee functional diversity and redundancy................................................................... 7
Factors structuring wild bee communities: Resources............................................... 9
Factors structuring wild bee communities: Environmental stressors........................ 11
Importance of pollination for Michigan highbush blueberry.................................... 15
Pest management in highbush blueberry................................................................... 18
Conservation strategies for wild bees........................................................................ 19
Thesis overview......................................................................................................... 21
CHAPTER 2: WILD BEE COMMUNITY RESPONSES TO PEST MANAGEMENT
AND LANDSCAPE COMPOSITION IN MICHIGAN HIGHBUSH BLUEBERRY......... 22
Introduction................................................................................................................ 23
Materials and Methods...............................................................................................27
Study sites...................................................................................................... 27
Bee collections............................................................................................... 29
Insecticide program risk................................................................................. 30
Landscape characterization............................................................................ 30
Functional traits............................................................................................. 31
Service providing bees................................................................................... 33
Data analysis.................................................................................................. 34
Results........................................................................................................................ 40
Bee collections............................................................................................... 40
Insecticide program risk................................................................................. 44
Landscape characterization............................................................................ 46
Bee responses to pest management and landscape........................................ 49
Comparison of social and solitary species..................................................... 59
Service providing bees................................................................................... 62
Discussion................................................................................................................. 63
CHAPTER 3: WILDFLOWER PLANTINGS INCREASE NESTING BY SOILNESTING BEES....................................................................................................................73
Introduction................................................................................................................ 74
vii

Materials and Methods............................................................................................... 79
Nesting suitability of blueberry farm habitats............................................... 79
Study sites.......................................................................................... 79
Emergence trapping........................................................................... 79
Microhabitat characteristics............................................................... 82
Soil characteristics............................................................................. 85
Bee communities collected in nets vs. emergence traps.................... 87
Bee identification............................................................................... 88
Suitability of new and mature wildflower restorations as nesting sites
for bees........................................................................................................... 88
Study sites.......................................................................................... 88
Emergence trapping........................................................................... 89
Microhabitat characteristics............................................................... 90
Bee communities collected in nets vs. emergence traps.................... 92
Results........................................................................................................................ 93
Nesting suitability of blueberry farm habitats............................................... 93
Microhabitat characteristics............................................................... 96
Soil characteristics............................................................................. 102
Bee communities collected in nets vs. emergence traps.................... 104
Suitability of new and mature wildflower restorations as
nesting sites for bees...................................................................................... 105
Microhabitat characteristics............................................................... 109
Bee communities collected in nets vs. emergence traps.................... 111
Discussion.................................................................................................................. 112
CHAPTER 4: CONCLUSIONS AND FUTURE DIRECTIONS......................................... 121
APPENDICES....................................................................................................................... 129
APPENDIX A: GPS COORDINATES OF FIELD SITES....................................... 130
APPENDIX B: COLLECTED WILD BEE SPECIES & ASSOCIATED
FUNCTIONAL TRAITS........................................................................................... 132
APPENDIX C: RECORD OF DEPOSITION OF VOUCHER SPECIMENS..........140
LITERATURE CITED.......................................................................................................... 145

viii

LIST OF TABLES

Table 2.1: Categories of habitat types used in landscape characterization (from Tuell,
2007)...................................................................................................................................... 31
Table 2.2: Traits used to calculate dispersion of functional traits of bees in the samples
collected in blueberry fields. Body size was condensed into a categorical variable, using
inter-tegular (IT) distance where available (Bartomeus et al., 2013).................................... 33
Table 2.3: Linear regressions of pooled average bee abundance and estimated richness
from 2004-2006 and 2013-2014 on the first two principal components from 2005
and 2014 landscape data at 300m, 500m, 1000m, and 1500m radii around sampled
sites........................................................................................................................................ 37
Table 2.4: Factor loadings for the first two principal components for landscape in 2005
and 2014. Loadings > 0.4 are shown in bold......................................................................... 49
Table 2.5. Slopes (β ± SE), coefficients of determination (R2), and p-values for simple
linear regressions of log-transformed wild bee abundance and Chao1 estimated species
richness on prior-year insecticide program risk for each year of collections at 13
blueberry farms...................................................................................................................... 51
Table 2.6. Slopes (β ± SE), degrees of freedom (df), and p-values for repeated measures
regressions of log-transformed wild bee abundance and Chao1 estimated species richness
on prior-year insecticide program risk for 2004-2006 collections, 2013-2014 collections,
and all five years of collections at 13 blueberry farms.......................................................... 51
Table 2.7: Model selection table for log-transformed wild bee abundance based on
repeated measures analysis on all combinations of insecticide program risk (IPR) and
the two first principal components for landscape composition (300 m radius). Models
are ranked according to Akaike’s information criterion (AIC), with the regression slopes,
AIC, ΔAIC, AIC weight, and relative variable importance based on Akaike weights
shown for the best competing models (ΔAIC ≤ 2) only........................................................ 53
Table 2.8: Best model for log-transformed wild bee abundance across five years of
sampling (2004-2006 and 2013-2014) at 13 blueberry farms............................................... 53
Table 2.9: Model selection table for Chao1 estimated richness for wild bees based on
repeated measures analysis on all combinations of insecticide program risk (IPR) and
the two first principal components for landscape composition (300 m radius). Models are
ranked according to Akaike’s information criterion (AIC), with the regression slopes for
model components, AIC, ΔAIC, AIC weight, and relative variable importance based on
Akaike weights shown for the best competing models (ΔAIC ≤ 2) only.............................. 53

ix

Table 2.10: Best model for Chao1 estimated wild bee richness across five years of
sampling (2004-2006 and 2013-2014) at 13 blueberry farms............................................... 54
Table 2.11: Model selection table for bias-corrected Shannon’s diversity of wild bees
based on repeated measures analysis on all combinations of insecticide program risk (IPR)
and the two first principal components for landscape composition (300 m radius). Models
are ranked according to Akaike’s information criterion (AIC), with the regression slopes
for model components, AIC, ΔAIC, AIC weight, and relative variable importance based
on Akaike weights shown for the best competing models (ΔAIC ≤ 2) only......................... 54
Table 2.12: Best model for bias-corrected Shannon’s diversity of wild bees collected
across five years of sampling (2004-2006 and 2013-2014) at 13 blueberry farms............... 54
Table 2.13: Model selection table for functional dispersion (FDis) of wild bees based
on repeated measures analysis on all combinations of insecticide program risk (IPR) and
the two principal components for landscape composition (300 m radius). Models are
ranked according to Akaike’s information criterion (AIC), with the regression slopes for
model components, AIC, ΔAIC, AIC weight, and relative variable importance based on
Akaike weights shown for the best competing models (ΔAIC ≤ 2) only.............................. 55
Table 2.14: Direction cosines and correlations of the three predictor variables with the
two-dimensional NMDS ordination of bee community composition.................................... 58
Table 3.1: Microhabitat characteristics measured in each emergence trap quadrat in
four habitat types on blueberry farms in 2013 and 2014....................................................... 82
Table 3.2: Microhabitat characteristics measured in each emergence trap quadrat in three
newly-established wildflower plantings, three mature wildflower plantings, and paired
unrestored controls in 2014.................................................................................................... 91
Table 3.3: Bee species collected in emergence traps from different habitat types on four
blueberry farms in 2013 and 2014. No. traps represents the number of individual traps
(and therefore the minimum number of possible nests) from which these species were
collected................................................................................................................................. 94
Table 3.4: Spearman’s rank correlation matrix of 11 microhabitat characteristics
measured in each 60cm2 emergence trap quadrat on four blueberry farms in 2013 and
2014 (n=645 traps)................................................................................................................ 97
Table 3.5: Factor loadings for the first two principal components from the microhabitat
characteristics analysis. Variables with factor loadings greater than 0.4 are in bold............ 98
Table 3.6: Model coefficients (β), z, p, and Akaike’s information criterion (AIC) values
for the competing zero-inflated models of bee abundance in traps, and the p-values from
likelihood ratio tests compared with a null model................................................................. 98

x

Table 3.7: Mean values by habitat for the microhabitat characteristics with large factor
loadings on PC1 and PC2. Values within a column followed by the same letter are not
significantly different (posthoc Dunn’s pairwise comparisons with Bonferroni correction,
α=0.05).................................................................................................................................. 98
Table 3.8: Model coefficients (β), z- and p-values for parameters, residual deviances,
degrees of freedom, and Akaike’s information criterion (AIC) values for the combined
and individual logistic regression models of microhabitat variables as predictors of bee
presence in emergence traps................................................................................................. 102
Table 3.9: Logistic regression model assessing the number of small cavities and percent
bare ground as predictors of the presence of bees in emergence traps................................. 102
Table 3.10: Soil physical properties in the crop field, grassy field margin, wildflower
restoration, and wooded areas on four blueberry farms in SW Michigan............................ 103
Table 3.11: Spearman’s rank correlation matrix of soil texture characteristics measured
on soil samples collected across different habitat types on four blueberry farms in 2013
(n=16).................................................................................................................................... 103
Table 3.12: Factor loadings on the first principal component for soil texture....................... 103
Table 3.13: Bee species collected in emergence traps from three newly-established and
three mature wildflower restorations and paired unrestored controls in 2014. No. traps
represents the number of individual traps (and therefore the minimum number of
possible nests) from which these species were collected...................................................... 106
Table 3.14: Microhabitat characteristics measured in emergence trap quadrats at three
new wildflower restorations, three mature wildflower restorations, and paired unrestored
control sites (n = 720 traps). Values within a column followed by the same letter are not
significantly different (posthoc Dunn’s pairwise comparisons with Bonferroni
correction, α=0.05)............................................................................................................... 108
Table 3.15: Spearman’s rank correlation matrix of microhabitat characteristics measured
in emergence trap quadrats at new wildflower restorations, mature wildflower
restorations, and paired unrestored control sites (n = 720 traps).......................................... 109
Table 3.16: Factor loadings on the first four principal components for microhabitat
characteristics in emergence trap quadrats in new wildflower restorations, mature
wildflower restorations, and their paired unrestored controls in 2014 (n=720 traps).
Factor loadings greater than 0.4 are in bold.......................................................................... 110
Table 3.17: Logistic regression model assessing the second and third principal
component axes as predictors of the presence of bees in emergence traps........................... 110
Table 3.18: Bee species most responsible for differentiating bee communities by
xi

collection method (SIMPER analysis).................................................................................. 112
Table A: GPS coordinates for the 15 blueberry fields in which bees were sampled
during bloom by J.K. Tuell between 2004-6 and by EM in 2013 and 2014........................ 130
Table B: GPS coordinates for the four blueberry farms sampled using emergence traps
in 2013 and 2014. Coordinates are for the wildflower restoration at each
site.......................................................................................................................................... 130
Table C: GPS coordinates for the new and mature wildflower restorations sampled using
emergence traps in 2013 and 2014.........................................................................................130
Table D: Wild bee species captured in pan traps and nets at 15 blueberry fields during
bloom by J.K. Tuell between 2004-2006 and EM in 2013 and 2014, and their associated
functional traits. Trait information was taken from Wolf & Ascher (2008), Bartomeus
et al. (2013), and Gibbs et al. (2011). Key to nest location: “Soil (R)” = species that
rent rather than excavate soil nests, and “Clepto” = cleptoparasitic species. Key to
sociality, “E” = eusocial, “F” = facultatively social, and “S” = solitary. Key to lecty
(diet specialization): “P” = polylectic and “O” = oligolectic. Body size was categorized
from inter-tegular distance, where available, or based on educated guesses (Bartomeus
et al., 2013). Key to body size: “S” = small, “M” = medium, and “L” = large (Table 2.2).
Service-providing bees are those that have been recorded collecting pollen from
Vaccinium corymbosum L. and close relatives (see reference column); a “no” in the
service-providing column means that the species has not been recorded visiting
Vaccinium flowers, or is a cleptoparasite or an oligolege on other types of
plants...................................................................................................................................... 132
Table E: List of voucher specimens. Each record represents one pinned adult specimen.
Species determinations (“Det”) were made by Dr. Jason Gibbs (JG) and the author
(EM)....................................................................................................................................... 140

xii

LIST OF FIGURES

Figure 2.1: Average percentage of bearing Michigan blueberry acres treated with each of
the eight conventional insecticides (solid line) and eight reduced-risk insecticides (dashed
line) reported by the National Agricultural Statistics Service (NASS 1991-2011).
Conventional insecticides included organophosphates, carbamates, and pyrethroids;
reduced-risk insecticides included biological insecticides, insect growth regulators, and
neonicotinoids........................................................................................................................ 25
Figure 2.2: Map of the study area showing the location of highbush blueberry sites
sampled in 2004-2006 (Tuell et al. 2009; Tuell & Isaacs 2010) and in 2013 and 2014
(this study)............................................................................................................................. 28
Figure 2.3: Species accumulation curve generated from bee community samples during
bloom in 2004-2006 and 2013-2014 at 15 blueberry farms. One sample represents the
community of bees collected in one day at one site (20 pan traps and 10 minutes of hand
netting). Dashed line represents the 95% confidence interval around the mean................... 41
Figure 2.4: Species accumulation curves for each year of sampling. One sample
represents the community of bees collected in one day at one of the 15 sites (20 pan
traps and 10 minutes of hand netting). At each site, there were two sampling rounds
during bloom in 2004 and 2006 (30 samples) and three sampling rounds in 2005, 2013,
and 2014 (45 samples). Error bars represent 95% confidence around the mean line for
each additional sample........................................................................................................... 42
Figure 2.5: Nonmetric multidimensional scaling ordination of bee communities collected
in pan traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014. Each point
represents one site in one year, with different colors for each of the five years. Ellipses
represent 95% confidence ellipses around the group mean for each year............................. 43
Figure 2.6: Nonmetric multidimensional scaling ordination of bee communities collected
in pan traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014. Each point
represents one site in one year. Black arrows show the trajectory of one site beginning in
2004 (circled points) and ending in 2014.............................................................................. 44
Figure 2.7: Boxplot of insecticide program risk in each of the five years prior to bee
collections.............................................................................................................................. 45
Figure 2.8: Increase in cumulative insecticide program risk (IPR) value over the 2012
growing season at the 13 conventionally managed blueberry fields. The average
cumulative IPR is represented by the black dashed line........................................................ 45
Figure 2.9: Biplot of first two principal components for landscapes surrounding blueberry
field locations sampled for bees at the 300 m scale in 2005.................................................. 48
xiii

Figure 2.10: Biplot of first two principal components for landscapes surrounding
blueberry field locations sampled for bees at the 300 m scale in 2014................................ 48
Figure 2.11: Simple linear regressions of log-transformed abundance of wild bees on
prior-year insecticide program risk (IPR) at the 13 blueberry farms that provided 5
years of application records. Dashed lines represent regression trend lines.......................... 52
Figure 2.12: Simple linear regressions of Chao1 estimated species richness of wild bees
on prior-year insecticide program risk (IPR) at the 13 blueberry farms that provided 5
years of application records. Dashed lines represent regression trend lines.......................... 52
Figure 2.13: Average untransformed abundance of wild bees per site captured over
five years of sampling during bloom at 13 blueberry farms, plotted against the average
prior-year insecticide program risk at those farms. Gray area represents 95% confidence
around the regression line. Two farms that did not provide all five years of spray records
were not included in the regression analysis and are not shown here................................... 55
Figure 2.14: Average untransformed abundance of wild bees per site captured over five
years of sampling during bloom at 13 blueberry farms, plotted against the second
principal component for landscape at 300 m (PC2) at those farms. Increasing PC2 is
associated with decreasing forest cover and increasing area of settlement (farm buildings,
houses, lawns, driveways, etc.). Grey area represents 95% confidence around the
regression line. Two farms that did not provide all five years of spray records were not
included in the regression analysis and are not shown here.................................................. 56
Figure 2.15: Repeated measures regression of prior-year insecticide program risk on
Chao1 estimated species richness of the wild bees captured over 5 years of sampling
during bloom at 13 blueberry farms. Grey area represents 95% confidence around the
regression line. Two farms that did not provide all five years of spray records were not
included in the regression analysis and are not shown here.................................................. 56
Figure 2.16: Average bias-corrected Shannon’s diversity of the wild bees captured over
five years of sampling during bloom at 13 blueberry farms, plotted against the average
prior-year insecticide program risk at those farms. Two farms that did not provide all
five years of spray records were not included in the regression analysis and are not shown
here......................................................................................................................................... 57
Figure 2.17: Nonmetric multidimensional scaling ordination of bee communities
collected in pan traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014.
Each point represents one site in one year. Black arrows show the trajectory of one site
beginning in 2004 (circled points) and ending in 2014. Red arrows show the projected
vectors of insecticide program risk (IPR) and the two first principal components for
landscape at the 300 m scale. Arrow lengths for IPR, PC1, and PC2 are scaled based on
correlation with the ordination configuration........................................................................ 55

xiv

Figure 2.18: Biplot of NMDS ordination of bee communities collected from 2004-2006
and 2013-2014, showing the projection of fitted vectors for the most abundant species
across all five years. Each colored dot represents one site in one year. Andrena imitatrix
and Lasioglossum cressonii, which overlapped strongly with A. miserabilis and L.
leucococum, respectively, are not shown............................................................................... 56
Figure 2.19: Nonmetric multidimensional scaling ordination of bee communities
collected in pan traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014.
Each colored dot represents one site in one year. Open black circles represent a single
species across all five years. Ellipses represent 95% confidence ellipses around the
group mean for solitary and social (including social, subsocial, and facultatively social)
bee species............................................................................................................................. 57
Figure 2.20: Repeated measures regressions of the average species richness of social
(blue line) and solitary (yellow line) wild bee species on the prior-year insecticide
program risk at 13 blueberry farms over five years of collections........................................ 59
Figure 3.1: Map of study sites sampled for soil nesting bees. Dark circles represent the
approximate locations of four blueberry farms sampled using emergence traps in four
different habitat types per farm in 2013 and 2014. Light rectangles represent newlyestablished wildflower restorations sampled in 2014. Light triangles represent mature
wildflower plantings sampled in 2014. Each restoration site had a paired unrestored
control site within 400 m (not marked).................................................................................. 76
Figure 3.2: Representative distribution of habitat types and sampling locations at one
of four blueberry sites sampled using emergence traps and netting in 2013 and 2014.
Imagery © 2015 GoogleTM..................................................................................................... 77
Figure 3.3: Scree plot of the percentage of variation explained by the principal
components generated from 11 microhabitat characteristics measured in emergence
traps in four different habitat types on blueberry farms in 2013 and 2014 (n=645 traps)..... 79
Figure 3.4: Scree plot of the percentage of variation explained by the 5 principal
components generated from the microhabitat data from emergence trap quadrats in new
wildflower restorations, mature wildflower restorations, and their paired unrestored
controls in 2014 (n=720 traps)............................................................................................... 87
Figure 3.5: Mean number of bees captured in emergence traps in different habitat types
on four blueberry farms in 2013 and 2014. Captured bee abundance does not differ
significantly among habitats marked with the same letter (posthoc Dunn’s pairwise
comparisons with Bonferroni correction, α=0.05)................................................................ 90
Figure 3.6: Biplot of first two principal components in the microhabitat PCA. Each dot
represents an individual emergence trap quadrat (n=645 traps). Ellipses represent 95%
confidence around the group mean for each habitat type...................................................... 96

xv

Figure 3.7: Non-metric multidimensional scaling of the community of bees captured in
emergence traps (etrap, black circles) and netted on open flowers (net, gray triangles) at
four blueberry farms in SW Michigan. Ellipses represent 95% confidence ellipses
around the group means......................................................................................................... 99
Figure 3.8: Mean number of bees collected per emergence trap in newly-established
wildflower plantings, mature wildflower plantings, and paired unrestored controls.
Locations with the same letter are not significantly different (posthoc Dunn’s pairwise
comparisons with Bonferroni correction, α=0.05)................................................................ 102
Figure 3.9: Boxplot of Chao1 estimated species richness of bees collected in emergence
traps in in newly-established wildflower plantings, mature wildflower plantings, and
paired unrestored controls. Locations with the same letter are not significantly different
(posthoc Dunn’s pairwise comparisons with Bonferroni correction, α=0.05)...................... 103
Figure 3.10: Non-metric multidimensional scaling of the community of bees captured
in emergence traps (etrap, black circles) and netted on open flowers (net, gray triangles)
in mature wildflower restorations, newly-established restorations, and paired unrestored
habitats for each restoration in SW Michigan. Ellipses represent 95% confidence ellipses
around the group means......................................................................................................... 106

xvi

KEY TO ABBREVIATIONS

BEF: Biodiversity-ecosystem functioning
FDis: Functional dispersion (Laliberté et al., 2010)
IPR: Insecticide program risk (Tuell & Isaacs, 2010)
perMANOVA: Permutational multivariate analysis of variance
NMDS: Non-metric multidimensional scaling

xvii

CHAPTER 1:

INTRODUCTION

1

Wild bees provide essential pollination services to pollinator-dependent crops and wild
plant populations (Klein et al., 2007; Kremen et al., 2007; Winfree et al., 2007). However, there
is increasing evidence that pollinator populations are threatened by a combination of stressors,
including habitat loss, degradation, and fragmentation, as well as an interacting combination of
pesticides, parasites, and disease (Biesmeijer et al., 2006; Potts et al., 2010; Cameron et al.,
2011; Goulson et al., 2015). Bee species are likely to differ in their sensitivity to environmental
disturbances based on life-history traits such as sociality, body size, and nesting substrate
(Larsen et al., 2005; Winfree et al., 2009; Williams et al., 2010; Brittain & Potts, 2011). These
differences in trait-mediated responses to disturbance may buffer communities against
environmental change (Elmqvist et al., 2003; Loreau et al., 2003; Fontaine et al., 2006).
Alternatively, environmental stressors may lead to non-random changes in the functional
diversity of bee communities (Brittain & Potts, 2011; Forrest et al., 2015). Understanding the
interplay of resources and stressors that structures bee populations and their distributions across
landscapes, as well as how species’ responses may be mediated by functional traits, will be
helpful in developing effective conservation strategies for these unmanaged insects (Winfree,
2010; Roulston & Goodell, 2011).

Biodiversity – ecosystem functioning
Human alteration of the natural environment, including activities such as deforestation,
agricultural intensification, and the movement of species among biomes, has led to rates of
species extinction estimated at 100 to 1,000 times pre-human levels in a wide range of taxonomic
groups (Pimm et al., 1995; Sala et al., 2000). This rapid species loss has generated intense
interest in research characterizing the relationship between biodiversity and ecosystem

2

functioning (BEF), which is the key element needed to predict the potential consequences of
biodiversity loss for ecosystem function (Chapin III et al., 2000; Loreau et al., 2001, 2002;
Hooper et al., 2005).
Interest in the nature or shape of the biodiversity-ecosystem functioning relationship
dates back to early work suggesting a connection between the structural complexity and stability
of ecosystems (MacArthur, 1955; Elton, 1958). The rich body of BEF research that emerged
over the decades following MacArthur’s original hypothesis was complicated by differing
measures and definitions of biodiversity – or complexity – and ecosystem function or stability,
which led to widely contradictory descriptions of the BEF relationship (May, 1973; Pimm, 1979,
1984; Naeem et al., 2009; Jax, 2010). Understanding the distinction between the different
variables used as proxies for complexity (species richness, interaction strength, connectance, and
species evenness) and stability (resilience, resistance, and variability) in different studies is
essential to drawing conclusions across studies (Pimm, 1984). Biodiversity-ecosystem
functioning assessments also require careful consideration of the specific ecosystem, reference
conditions, and spatial and temporal scales of observation in each study (Loreau et al., 2001; Jax,
2010).
One area of biodiversity-ecosystem functioning research that has received increasing
attention over the past 20 years is the relationship between the functional diversity of organisms
– rather than diversity of taxonomic identities – and ecosystem stability or function. Functional
diversity is governed by ecological niche breadth (specialization) and niche differentiation
(complementarity), both of which interact to allow competing species to coexist given limited
resources (Blüthgen & Klein, 2011). The redundancy of ecological niches or functions can also
be important in protecting ecosystems against loss of function as species are lost (Naeem, 1998;

3

Rosenfeld, 2002; Winfree & Kremen, 2009). Assessing loss of biodiversity through the lens of
species’ functional traits and niches allows for a mechanistic understanding of how species loss
may affect ecosystem function (Rosenfeld, 2002).

Importance of pollination in natural and managed systems
Pollination is one example of a crucial ecosystem process that may be affected by
biodiversity loss (Kremen, 2005; Larsen et al., 2005; Tscharntke et al., 2005; Winfree &
Kremen, 2009; Blüthgen & Klein, 2011). Pollination, or the transfer of pollen from the male to
the female parts of a plant to facilitate reproduction, is a critically important process in both
natural and managed systems. Without pollination, many plant species would not set seed or
fruit, leaving them unable to reproduce or to provide food resources for humans and other
consumers (McGregor, 1976; Free, 1993; Klein et al., 2007). While many animal-pollinated
species have mixed mating systems that allow individuals to self-pollinate in the absence of
pollinators, in the long term, these species also require outcrossing to maintain gene flow
between members of the species.
Of the more than 350,000 species of angiosperms found worldwide, the majority are
pollinated by insects and other animals, with a small minority receiving pollination via abiotic
vectors such as wind and water (Ollerton et al., 2011). It has been estimated that over 80% of
flowering plant species and 75% of the leading global food crops depend on animal pollinators,
primarily bees (Klein et al., 2007; Ollerton et al., 2011). Globally, the managed European honey
bee (Apis mellifera L.) is the dominant crop pollinator (Crane, 1990; Williams, 1994; Delaplane
et al., 2000; Moritz et al., 2005). However, a recent meta-analysis determined that wild bee
visitation is more strongly correlated with fruit set than honey bee visitation in many cultivated

4

crops (Garibaldi et al., 2013), indicating that honey bees cannot fully substitute for diverse wild
bee pollinators, nor do they maximize pollination service delivery in many systems. In addition,
the cultivated area of pollinator-dependent crops has increased disproportionately relative to
other agricultural crops over the past half-century, increasing demand for pollination services at a
pace that well outstrips the rate of increase in the global stock of managed honey bees (Aizen et
al., 2008a; Aizen & Harder, 2009). Wild bees and the pollination services they provide are
therefore increasingly essential to the output and stability of global crop production; however,
their populations are threatened by agricultural intensification and other environmental stressors
(Kremen et al., 2002; Potts et al., 2010; Goulson et al., 2015).

Overview of pollinator declines
Large-scale land use intensification over the past century has resulted in widespread
declines in farmland biodiversity (Matson et al., 1997; Krebs et al., 1999; Tscharntke et al.,
2005; Gonthier et al., 2014). Declines in global wild bee abundance and species richness
associated with agricultural intensification may threaten the provision of pollination services to
wild and domesticated plant species (Allen-Wardell et al., 1998; Kearns et al., 1998; Kremen et
al., 2002; Biesmeijer et al., 2006; Potts et al., 2010). It is challenging to document declines in
pollinator assemblages due to their high annual species turnover and the labor costs associated
with intensive monitoring programs (Cane & Tepedino, 2001; Roubik, 2001; Williams et al.,
2001; Winfree, 2010). However, historical collection records have provided evidence for
widespread bee declines (pre- vs post-1980s) in the United Kingdom (Biesmeijer et al., 2006),
particularly bumble bees (Bombus spp.) (Goulson et al., 2008). Museum records have also been
used to test for evidence of bumble bee declines in North America (Colla & Packer, 2008; Grixti

5

et al., 2009; Cameron et al., 2011; Bartomeus et al., 2013). However, Bartomeus et al. (2013)
found only weak declines in aggregate native bee richness for northeastern US species outside of
the genus Bombus, with some species exhibiting significant declines in relative abundance over
time but others exhibiting significant increases. Ecological traits associated with declining
species included small dietary and phenological breadth and large body size (Bartomeus et al.,
2013). There is also indirect evidence for declines from numerous studies indicating reduced bee
abundance and diversity in intensive agricultural environments compared with natural or seminatural areas (Kremen et al., 2002; Winfree et al., 2009; Potts et al., 2010).
Agricultural intensification occurs along a range of spatial scales: at the local scale,
intensification leads to decreases in within-farm crop diversity and increases in chemical or other
disturbance inputs, while at the landscape scale, intensification leads to increasingly homogenous
cropped landscapes with reductions in natural and semi-natural habitat (Benton et al., 2003;
Tscharntke et al., 2005; Kennedy et al., 2013). For bees, the result is both a local- and landscapelevel loss of food and shelter resources and increasing exposure to stressors, including parasites,
pathogens, diseases, and the use of pesticides, particularly insecticides, in agricultural and urban
environments (Kearns et al., 1998; Steffan-Dewenter et al., 2002; Tscharntke et al., 2005;
Carvell et al., 2006a; Hendrickx et al., 2007; Potts et al., 2010; Roulston & Goodell, 2011;
Vanbergen & Insect Pollinators Initiative, 2013; Goulson et al., 2015). Many of these possible
drivers of bee declines are interrelated and synergistic; for example, a bee with a weakened
immune system due to sublethal pesticide exposure might be more susceptible to pathogen
infection or its energetic consequences (Alaux et al., 2010a; Pettis et al., 2012; Goulson et al.,
2015). More broadly, habitat loss and fragmentation concentrates bees on smaller resource

6

patches, which may lead to higher rates of pathogen and disease transmission and greater
chances of exposure to agricultural chemicals.

Bee communities and ecological diversity
The bees (Hymenoptera: Apoidea) are a large and diverse insect group, with an estimated
20,000 species worldwide, including more than 400 species in Michigan (Michener, 1979;
Michener, 2000; JG, unpublished). Bees are functionally and ecologically diverse, with wide
variation in life history traits such as sociality, body size, trophic specialization, and nesting
location (Michener, 2000; Williams et al., 2010).
Wild bee communities are variable in space and time, with high species turnover across
and within years (Williams et al., 2001; Petanidou et al., 2008; Mandelik et al., 2012; Sydenham
et al., 2014). Communities are generally comprised of diverse local fauna, often with many rare
species present, and only a small proportion of species remain from year to year (Williams et al.,
2001). Within years, bee communities also move to track variation in resources across
landscapes, so a locally species-rich community in early spring may have only a few species
present in the same area in late summer (Mandelik et al., 2012; Sydenham et al., 2014).

Bee functional diversity and redundancy
Wild bee diversity is associated with increased fruit and seed set of insect-pollinated wild
and domesticated plant species, particularly when measured in terms of functional diversity
instead of species richness (Klein et al., 2003; Fontaine et al., 2006; Hoehn et al., 2008; Albrecht
et al., 2012; Rogers et al., 2014). The functional diversity of pollinators – specifically, diversity
in length of mouthparts - has been experimentally shown to increase the persistence of plant

7

communities by increasing plant reproductive success (Fontaine et al., 2006). Conversely, the
functional diversity of plant communities, such as complementary differences in flowering
phenology or the nutritional content of nectar/pollen resources, is thought to increase the
persistence of pollinator communities (Blüthgen & Klein, 2011).
In addition to diversity, functional redundancy in bee communities may buffer plantpollinator networks, or the interaction networks between communities of plants and associated
pollinators, against the consequences of species loss (Blüthgen & Klein, 2011). Plant-pollinator
networks are generally highly nested, with specialist bee species visiting subsets of the plant
species visited by generalist bees (Bascompte et al., 2003; Pawar, 2014; Rohr et al., 2014). Put
another way, specialization in plant-pollinator networks is highly asymmetric; oligolectic bees,
or bees with narrow diet breadth, are much more likely to visit generalist plants that are visited
by many species of bees than plant species visited by only a few species (Vázquez & Aizen,
2004). This asymmetrical specialization and associated nestedness leads to redundancy in
pollination function for the plant species visited by both specialist and generalist bees, creating
stability in the network that allows rarer species of both plants and bees to persist (Bascompte et
al., 2003; Pawar, 2014; Rohr et al., 2014). Theoretical work has shown that nestedness can
increase the robustness and persistence of plant-pollinator networks to environmental disturbance
(Memmott et al., 2004), with the caveat that there may be ecological “tipping points” after which
nested communities exhibit sudden collapse (Lever et al., 2014).
The consequences of bee species loss for plant-pollinator networks and crop pollination
also depend on the order of species loss, which is a function of the susceptibility of different bee
species to extinction pressures and their relative functional contributions (Larsen et al., 2005). If
the most functionally important bees are also the most at risk of extinction, plant-pollinator

8

networks would be expected to lose function rapidly as species are lost. Larsen et al. (2005)
hypothesized that the order of species loss – and associated loss of function – in bee communities
will be highly non-random. Large-bodied bees, often with higher pollination efficiencies (e.g.
Bombus spp.), are more susceptible to extinction pressures (Larsen et al., 2005). If this
hypothesis is supported, loss of species may be quickly devastating for ecosystem function.
However, the high annual and seasonal species turnover and variation in species interactions in
plant-pollinator networks, as well the opportunistic nature of many plant-pollinator interactions,
may mean that these networks are less susceptible to disturbance than these dire estimates would
suggest (Petanidou et al., 2008).

Factors structuring wild bee communities: Resources
The major factors governing the organization of bee community assemblages across
spatiotemporal scales are the availability of floral and nesting resources and the nature and
strength of environmental stressors (Potts et al., 2003, 2005; Williams & Kremen, 2007;
Roulston & Goodell, 2011). Bees feed their offspring with a mixture of pollen, nectar, and
(rarely) plant oils (Michener, 2000). The quantity and quality of floral resources are generally
accepted as the main drivers of bee community structure, abundance and richness (Potts et al.,
2003; Hines & Hendrix, 2005), and their distribution can affect bees’ reproductive outputs
(Williams & Kremen, 2007). In the absence of sufficiently abundant and diverse floral resources,
poor nutrition can lead to reduced larval development, colony growth, and stress resistance, as
well as reduced fecundity, immunocompetence, and longevity in adult bees (Haydak, 1970;
Hoover et al., 2006; Alaux et al., 2010b; Brodschneider & Crailsheim, 2010; Huang, 2012;
DiPasquale et al., 2013). In considering the relationship between floral resource availability and

9

bee community structure, abundance, and diversity, it is important to note that bee communities
are affected by the availability of floral resources at multiple time points during their life cycle,
and that the availability of resources during the period of offspring production in one year may
be unconnected to the resources available for the emerging adults the following year (Tepedino
& Stanton, 1981).
Nesting resources also play an important role in organizing bee communities, though
there are limited data on bees’ preferences for specific substrate characteristics (Potts et al.,
2003, 2005; but see Cane, 1991). Approximately two-thirds of bees nest in the soil, with the
remaining third nesting above ground, generally in pre-existing cavities in dead wood or in pithy
stems or canes (Cane, 1991; Michener, 2000). Potts et al. (2003) found that floral characteristics
were the most important predictors of bee community structure in a Mediterranean habitat, but
the diversity of nesting habitats explained only 5-10% of the community structure. A later study
in the same Mediterranean environment found that the availability of nesting resources explained
40-60% of the variation in species-abundance pattern, in particular the availability of bare
ground and suitable nesting cavities (Potts et al., 2005). There remain significant knowledge
gaps in understanding nest site preferences, particularly for soil-nesting bees. Only a few studies
have examined the nesting requirements and preferences of bees, and for a small number of
species (Michener et al., 1958; Osgood, 1972; Cane, 1991; Potts & Willmer, 1997; Svensson et
al., 2000; Potts et al., 2005; Kim et al., 2006; Sardiñas & Kremen, 2014). These gaps in our
understanding also undermine the predictive capacity of landscape models of pollinator
abundance, which could be important tools for developing landscape-scale pollinator
conservation strategies (e.g. Lonsdorf et al., 2009).

10

Factors structuring wild bee communities: Environmental stressors
Along with resource distributions, there are several interacting environmental stressors
that may act as filters on local and regional species pools (Keddy, 1992; Kraft et al., 2014;
Sydenham et al., 2014). Environmental “filters” are local or regional environmental conditions
that place restrictions on the ability of species to persist in those environments based on different
species’ traits, which thus determine the species - and trait - composition of communities living
in those environments (Keddy, 1992). In one example of how environmental filtering might
determine community assembly, a wetland-adapted plant species is unlikely to persist in dry site
conditions, even if the species exists within dispersal range of the dry site. With enough
information on species’ functional traits and environmental conditions, an environmental
filtering framework could be used to predict the species composition of different sites based on a
given set of environmental conditions and the range of species’ trait values for the functional
traits that determine species’ responses to those environmental conditions (Keddy, 1992).
Environmental stressors potentially affecting wild bees include exposure to pesticides
(Desneux et al., 2007; Brittain et al., 2010b; Tuell & Isaacs, 2010a); infection with diseases,
pathogens, and parasites (Colla et al., 2006; Otterstatter & Thomson, 2008; Singh et al., 2010;
Cameron et al., 2011); as well as the possible effects of climate change (Hegland et al., 2009;
Schweiger et al., 2010; Bartomeus et al., 2011), invasion of alien plant species (Cox & Elmqvist,
2000; Aizen et al., 2008b; Kleijn & Raemakers, 2008; Stout & Morales, 2009), and competition
with managed bees (Goulson, 2003a). These stressors may also interact with the distribution of
resources at different scales across the landscape to determine community assembly at the local
scale (Sydenham et al., 2014). Individual bee species are able to persist in local environments
only if their functional traits are compatible with the local conditions; that is, if local resources

11

are sufficiently available and accessible, and if species are not “filtered out” by the set of local
environmental stressors.
Bees can be exposed to a diverse array of pesticides across the landscape (Chauzat et al.,
2006; Johnson et al., 2010; Mullin et al., 2010; Wu et al., 2011; Krupke et al., 2012). Pesticide
risk to bees is a function of the route and length of exposure and the toxicity of the chemical.
Exposure to a pesticide, whether by contact or ingestion, can be through direct contact from an
application or indirect through contact with foliar residues or spray drift, or via ingestion of
systemic compounds expressed in nectar or pollen, etc. (Thompson, 2001; Desneux et al., 2007;
Krupke et al., 2012). Chronic exposure can also affect bees, for example if developing brood are
fed contaminated pollen or nectar, or if foraging bees are routinely exposed to low levels of
residues or spray drift across the landscape. All of these types of exposure can have deleterious
effects on managed and wild bees.
Pesticide toxicity to bees is most commonly measured as the contact LD50 for honey bees,
or the dose that kills 50% of a given population of adults upon external contact exposure, which
is required for the registration of a pesticide by the US Environmental Protection Agency.
However, honey bee responses to different chemicals may not always be representative of the
risk to other bee species (Arena & Sgolastra, 2014), and only a handful of chemicals have been
tested for their toxicity to other bee species, particularly solitary species (Helson et al., 1994;
Thompson, 2001; Morandin et al., 2005; Scott-Dupree et al., 2009; Cresswell et al., 2012;
Biddinger et al., 2013).
Bees exhibit a wide range of physiological and behavioral responses to pesticides, which
can differ significantly among different taxonomic groups based on life history and ecological
traits (Thompson, 2001, 2003; Desneux et al., 2007; Brittain & Potts, 2011). For example, small-

12

bodied bees may be more susceptible to insecticide exposure, since their higher surface area to
volume ratio may increase contact absorption relative to larger bodied bees (Brittain & Potts,
2011). Other life-history traits, such as sociality, can be less straightforward in their effects on
susceptibility to pesticides, but these types of trait-based responses to pesticide exposure could
lead to non-random changes in bee community structure due to disturbance by pesticides
(Brittain & Potts, 2011). A few studies have found community-level effects of pesticides on wild
bees in agricultural landscapes, with declines in wild bee species richness reported in two
blueberry systems and one diversified European agricultural landscape (Kevan, 1977; Brittain et
al., 2010b; Tuell & Isaacs, 2010a).
In many agricultural systems, it is common to apply a variety of insecticides through the
season to control different insect pests, as well as fungicides to control diseases; often these are
applied as tank mixes of insecticides and fungicides. While fungicides have generally been
considered safe to bees, several studies have shown that some fungicides can exhibit synergistic
toxicity with certain insecticides (Pilling & Jepson, 1993; Thompson & Wilkins, 2003; Iwasa et
al., 2004; Biddinger et al., 2013). For example, ergosterol biosynthesis inhibiting (EBI)
fungicides have been found to synergize the toxicity of the pyrethroid class and some members
of the neonicotinoid class of insecticides by interfering with the cytochrome P450 detoxification
pathway in honey bees (Pilling & Jepson, 1993; Thompson & Wilkins, 2003; Iwasa et al., 2004).
Synergistic interactions may also occur with varroacides applied for in-hive mite treatments for
honey bees, or with certain combinations of insecticides that bees may be exposed to while
foraging across a diverse landscape (Brittain et al., 2010b; Johnson et al., 2010; Gill et al., 2012;
Goulson et al., 2015). However, few studies have measured synergistic interactions of chemicals
at field-realistic levels, particularly for wild bee species (Gill et al., 2012; Biddinger et al., 2013).

13

Many diseases, pathogens, and parasites can be spread within and between populations of
wild and managed bees (Colla et al., 2006; Otterstatter & Thomson, 2008; Singh et al., 2010;
Cameron et al., 2011). The spillover of pathogens from commercially managed bumble bees to
wild populations has been implicated in the disappearance of a western North American bumble
bee species (Bombus franklini) and the decline of several other species in the same subgenus
(Colla et al., 2006; Otterstatter & Thomson, 2008; Colla & Ratti, 2010; Cameron et al., 2011;
Szabo et al., 2012). Cameron et al. (2011) found that declining populations of Bombus have
significantly higher infection levels of the microsporidian pathogen Nosema bombi, which can
reduce individual fitness and longevity as well as colony growth (Otti & Schmid-Hempel, 2007,
2008). While honey bee pathogens have been fairly well-studied over the last 50 years, the scope
of transmission of viruses and parasites between managed honey bees and wild bees has only
begun to be recognized (Fürst et al., 2014; Ravoet et al., 2014). Many diseases previously
thought to be restricted to honey bees, such as deformed wing virus and the Varroa-transmitted
Macula-like virus, have now been discovered in populations of wild bees, including several
species of solitary bees (Fürst et al., 2014; Ravoet et al., 2014). These new findings suggest that
the movement of managed honey bee colonies for commercial pollination may spread pathogens
to wild bee populations, which may in turn act as reservoirs of honey bee pathogens (Ravoet et
al., 2014).
Climate change is an additional factor that may lead to extinctions of bee species if they
or their host plants are unable to shift their ranges to accommodate their thermal tolerance, or if
range shifts cause temporal or spatial mismatches in plant-pollinator interactions (Hegland et al.,
2009; Schweiger et al., 2010; Winfree, 2010). Using historical collection data from northeastern
North America, Bartomeus et al. (2011) showed that plant and pollinator phenological advances

14

have occurred at similar rates in response to long-term temperature increases for ten common
generalist bee species. Similar analytical methods using museum data are not possible for rare
specialist species, for whom phenological mismatches with host plants may be more devastating
(Bartomeus et al., 2011). Diversity of bee communities may help to buffer plant-pollinator
networks against shifts in flowering phenology (Bartomeus et al., 2013). However, increasing
temperatures and more frequent droughts may reduce flower production and nectar volume in
flowering plants, leading to nutritional stress for pollinators regardless of range or phenology
(Carroll et al., 2001; Fang et al., 2010; Minckley et al., 2013). In addition, the climate is
predicted to become more variable, with more erratic rainfall and larger fluctuations in weather
conditions on a short time scale (Francis & Vavrus, 2012; Thornton et al., 2014). This increase in
variability could be more problematic for bees and spring-flowering crops than gradual changes
in mean climate conditions, with greater chance of floods, drought, and variable spring weather
during key foraging periods for early spring bees (Katz & Brown, 1992; Tuell & Isaacs, 2010b;
Francis & Vavrus, 2012).
None of these stressors act in isolation, and the combined effects of multiple stressors are
likely to be more harmful than any single source of environmental stress (Goulson et al., 2015).
Complex causes of pollinator declines mean that strategies for conservation and restoration of
bee communities need to address these complexities.

Importance of pollination for Michigan highbush blueberry
The highbush blueberry (Vaccinium corymbosum L.) agroecosystem in southwest
Michigan is an ideal system in which to examine the complex interacting stressors that bees face
in agricultural environments. The landscape is a heterogeneous mix of agriculture, forest, settled

15

areas, and some fallow and ruderal areas, but is chiefly dominated by commercial fruit
production, with the associated characteristics of agricultural intensification: pesticide use and
relatively scarce floral resources when the crop is not in bloom. In addition, the main crop is
dependent on pollinators for marketable yields, creating an incentive for growers to be interested
in the conservation of wild bees.
Michigan is the nation’s leading producer of cultivated blueberries, with 114 million
pounds of blueberries valued at $121 million dollars, produced on 19,000 acres in 2013 (NASS,
2014). Highbush blueberry, the main blueberry species grown in Michigan, is dependent on
insect-mediated pollination for economically viable yields (Merrill, 1936; Schaub & Baver,
1942; Meader & Darrow, 1947; Free, 1993). At least 80% of flowers must set fruit to achieve a
commercial-scale blueberry crop (Merrill, 1936). Though cultivated varieties of V. corymbosum
exhibit varying degrees of self-compatibility, cross-pollination generally produces earlier
ripening berries, increased berry size, and higher seed and fruit set (Schaub & Baver, 1942;
Meader & Darrow, 1947; Dorr & Martin, 1966; Brewer et al., 1969; Krebs & Hancock, 1990;
Lang & Danka, 1991; VanderKloet, 1991; Hokanson & Hancock, 2000; Chavez & Lyrene, 2009;
MacKenzie, 2009).
Blueberry growers typically rent commercial honey bee hives to fulfill pollination needs
(Brewer et al., 1969; Isaacs & Kirk, 2010). Isaacs and Kirk (2010) estimated that honey bees
(Apis mellifera L.) provide 88% of the yield increase due to pollination in the Michigan highbush
blueberry system, with wild bees providing the remaining 12%. However, wild bees are the
dominant pollinators in small, isolated fields, providing a level of pollination comparable to what
stocked honey bees provide in large commercial fields. Over 100 wild bee species have been

16

recorded in Michigan highbush blueberry fields during bloom, including at least 10 species
exhibiting high abundance and/or fidelity to blueberry flowers (Tuell et al., 2009).
Diverse wild bee communities, with their diversity of ecological and life-history traits,
may ensure the stability of pollination service delivery to highbush blueberry (Rogers et al.,
2014). Many wild bees, particularly those species capable of vibratile pollination (hereafter
“buzz pollination”), exhibit high pollination efficiencies on Vaccinium flowers (Buchmann,
1983; Cane & Payne, 1993; Sampson & Cane, 2000; Stubbs & Drummond, 2001; Javorek et al.,
2002; Dedej & Delaplane, 2003; Desjardins & De Oliveira, 2006; Rogers et al., 2013). In the
closely related lowbush blueberry (Vaccinium angustifolium Aiton), Bombus spp. queens and
Andrena spp. were found to pollinate 3-6x more flowers than honey bees in the same amount of
time, and deposit 4x the amount of pollen in a single visit (Javorek et al. 2002). In British
Columbia, blueberry weight in six highbush blueberry fields was correlated with the abundance
of bumble bees (Bombus spp.), but not the abundance of honey bees or other wild bees (Ratti et
al., 2012). Furthermore, large-bodied wild bees, such as Bombus spp., are capable of foraging in
the suboptimal weather conditions typical of spring blueberry bloom (Heinrich, 2004; Tuell &
Isaacs, 2010b). Rogers et al. (2014) found that while honey bees were three times less abundant
in blueberry fields in inclement weather, overall wild bee density did not differ between
inclement and optimal weather conditions due to the stable abundances of large-bodied bee
genera such as Bombus, Habropoda, and Xylocopa. The biodiversity insurance that these wild
bee communities provide may help buffer the blueberry agroecosystem against fluctuations in
honey bee pollination supply (Winfree et al., 2007; Rogers et al., 2014).

17

Pest management in highbush blueberry
Though blueberry is highly dependent on bees for pollination during bloom, many fields
are managed intensively to control pests and diseases throughout the growing season, with
potential negative consequences for the wild bees living in and around fields after bloom. While
short-lived bees that are tightly linked with blueberry bloom are unlikely to be affected by postbloom insecticides, the sprays may have fitness consequences for bees with longer life cycles,
including bumble bees (Bombus spp.). A 3-year pan trapping study found that wild bee
abundance and species richness in blueberry fields declined with increasing insecticide use in 2
of 3 years, suggesting that season-long pest management program intensity can affect the bee
community emerging during bloom the following year (Tuell & Isaacs, 2010a). Similarly, Kevan
et al. (1997) found that the diversity and abundance of wild bees in lowbush blueberry fields
fitted a log-normal model in fields unaffected by pesticide applications to the surrounding
forests, but departed from the model in fields subject to pesticide stress.
For the past ten years, blueberry production in Michigan has trended toward greater
adoption of integrated pest management (IPM) strategies, including reductions in the quantity
and toxicity of insecticides applied per season (NASS, 1991-2011). The use of broad-spectrum
insecticides, such as azinphos-methyl, malathion, and carbaryl, declined strongly from 20012011 in favor of reduced-risk insecticides such as acetamiprid and methoxyfenozide. However,
the recent arrival of spotted-wing drosophila (Drosophila suzukii Matsumura, Diptera:
Drosophilidae), an invasive pest that can cause major economic damage to stone and small fruits,
threatens the continued viability of IPM strategies in blueberry and other susceptible fruit crops
(Bolda et al., 2010; Hauser, 2011; Lee et al., 2011; Walsh et al., 2011). Estimates for Michigan
indicate 2012 losses due to D. suzukii at nearly $27 million (R. Isaacs, unpublished), and the eFly

18

SWD Working Group has estimated $207 million losses from SWD in the eastern US alone
(eFly, 2012). This invasive pest, with its short generation time, high dispersal ability, and lack of
natural enemies, has led to Michigan blueberry growers applying broad-spectrum insecticides in
a prophylactic manner to control this insect prior to harvest, a trend that may be highly
deleterious to the bees and natural enemies active in and around crop fields after bloom.

Conservation strategies for wild bees
Given the importance of wild bees for pollination of wild and cultivated plants and the
growing indications of declining wild bee populations, conservation strategies are needed to
support, restore, and enhance diverse and abundant wild bee populations, particularly in
intensified agricultural landscapes where resources are scarce and pollination services are most
needed (Kremen et al., 2002; Winfree, 2010). Multiscale conservation strategies for bees can
include species-targeted approaches, such as listing individual species as threatened or
endangered to receive federal protections, or habitat-targeted approaches, such as preserving
natural areas or restoring lost, degraded, or fragmented habitats (Edwards, 1996; Murray et al.,
2009c; Winfree, 2010; Gonthier et al., 2014). Cane (2001) defines habitat for bees as “minimally
consist[ing] of rewarding patches of floral resources plus suitable nesting sites, all within flight
range of each other.” The scale of the necessary habitat patchwork varies based on the nutritional
requirements and flight radius of different bee species. Successful habitat restoration for bee
conservation, therefore, will incorporate provision of both floral and nesting resources at
multiple spatial scales to support diverse bee communities and the pollination services they
provide (Potts et al., 2005).

19

There is considerable evidence to show that restoring diverse floral resources to resourcepoor environments can increase local bee abundance and diversity (Kells et al., 2001; Carvell et
al., 2007; Pywell et al., 2007; Blaauw & Isaacs, 2014a, 2014b). There are many possible
strategies for introducing additional floral resources into landscapes, including the planting of
flowering cover crops, sequentially blooming agricultural crops, spring-flowering shrubs and
trees, and/or annual or perennial forb mixes. Of these, the easiest to incorporate in the marginal
areas around perennial fruit plantings is likely to be perennial forb mixes, which can be tailored
to fit site conditions and can provide floral resources for years following the initial planting, but
do not necessarily represent a permanent investment of land (as flowering trees and shrubs
might).
These perennial wildflower plantings can provide diverse, season-long nectar and pollen
resources for adult bees and their developing larvae (Carreck & Williams, 2002; Carvell et al.,
2006b; Pywell et al., 2007; Blaauw & Isaacs, 2014b; Williams et al., 2015). Perennial plantings
may also supply nesting and hibernation sites for ground-nesting and stem-nesting bees (Carvell
et al., 2007). Using molecular markers, Wood et al. (2015) found that floral restorations
significantly increased the nest density of four bumble bee species (Bombus spp.) on farms in the
UK. However, most bee species are solitary and their nest density cannot be assessed with
molecular tests based on relatedness of collected specimens. One technique that has shown
promise for evaluating nest density of these species is soil emergence traps, which can be used to
capture emerging bees over several days or over the course of a growing season (Sardiñas &
Kremen, 2014). Stem and cavity-nesting species can be relatively easily monitored in different
landscapes with installations of wood blocks, cardboard tubes, or natural reeds (Krombein, 1967;
Tscharntke et al., 1998a; Buschini, 2006; Cane et al., 2007). These aboveground-nesting species

20

may be more nest site limited than soil-nesting species in highly simplified agricultural
landscapes (Williams et al., 2010), so conservation schemes may also need to incorporate
provision of additional nesting resources for these bees.
Establishing forb-rich wildflower patches, strips, or meadows near pollinator-dependent
crops, such as blueberry, can result in net economic benefits for growers by increasing
pollination and yield in those crops (Carvalheiro et al., 2012; Blaauw & Isaacs, 2014b). Using
crop yield scenarios estimated from berry weight data, Blaauw and Isaacs (2014b) found that the
increased pollination service spillover from wildflower plantings sown on marginal land adjacent
to blueberry could generate a yield benefit exceeding the costs of planting establishment and
maintenance within five years. These ecosystem service benefits, along with financial incentive
programs that support beneficial insect conservation, can help to encourage grower adoption of
these conservation practices (Kennedy et al., 2013; Blaauw & Isaacs, 2014b).

Thesis overview
The main goal of my research is to determine the most effective strategies for conserving
and enhancing wild bee populations in the Michigan blueberry agroecosystem. Within this, my
specific research objectives were:
1) to determine how pest management at the field scale and landscape composition at a
broader scale have affected wild bee communities in highbush blueberry fields before and
after the arrival of D. suzukii.
2) to determine which habitats in Michigan farm landscapes are most suitable for soilnesting bees, and whether bees nest preferentially in wildflower plantings compared with
nearby unrestored habitats.

21

CHAPTER 2:

WILD BEE COMMUNITY RESPONSES
TO PEST MANAGEMENT AND LOCAL LANDSCAPE COMPOSITION
IN MICHIGAN HIGHBUSH BLUEBERRY

22

Introduction
An estimated 80% of wild plant species and 75% of the leading global food crops depend
on animal pollinators, primarily bees (Klein et al., 2007). Declines in global wild bee abundance
and richness, especially in intensively managed agricultural areas, may be threatening the
provision of pollination services to wild and domesticated plant species (Allen-Wardell et al.,
1998; Kremen et al., 2002; Biesmeijer et al., 2006; Potts et al., 2010). A number of possible
causes for these declines have been hypothesized, including habitat loss, fragmentation, and
degradation; interacting effects of parasites, pathogens, and diseases; and the use of pesticides,
particularly insecticides, in agricultural and urban environments (Kearns et al., 1998; Tscharntke
et al., 2005; Roulston & Goodell, 2011; Goulson et al., 2015). Understanding how these factors
interact with resource availability and landscape composition to determine bee community
assembly will be important in developing effective wild bee conservation strategies, particularly
in agricultural areas that require robust wild bee populations to support pollination of fruits and
vegetables. In this study, I investigated the interactive effects of landscape composition and
insecticide use on wild bee communities in the intensively managed highbush blueberry
agroecosystem of southwest Michigan.
Highbush blueberry (Vaccinium corymbosum L.) requires insect pollination for
economically viable yields (Merrill, 1936; Schaub & Baver, 1942; Meader & Darrow, 1947;
Free, 1993; Isaacs & Kirk, 2010). While most Michigan blueberry growers rely on rented honey
bee (Apis mellifera L.) colonies for pollination, the high winter losses of managed honey bee
colonies across North America over the last 20 years, combined with the rising costs of hive
management, have increased hive rental prices and weakened remaining colonies (Rucker et al.,
2012). The rising costs and lower returns associated with honey bee rental underscore the

23

importance of supporting wild pollinator populations to ensure the resilience of pollination
services and crop yields (Allen-Wardell et al., 1998; Winfree et al., 2007).
Over 100 wild bee species have been recorded in Michigan highbush blueberry
(Vaccinium corymbosum L.) fields during bloom, including at least 10 species exhibiting high
abundance and/or fidelity to blueberry flowers (Tuell et al., 2009). However, management of
insect pests following bloom often necessitates several insecticide applications per season. While
short-lived bee species that are temporally linked with blueberry bloom may not be affected by
post-bloom insecticides, these sprays may have fitness consequences for bees with longer life
cycles. Several bumble bee species (Bombus spp.), for example, are efficient blueberry
pollinators that emerge prior to bloom and continue to forage and reproduce throughout the
growing season, potentially placing them at risk of exposure to the insecticide use during berry
ripening and harvest. Within this system, season-long pest management program intensity has
been negatively correlated with wild bee abundance and species richness (Tuell & Isaacs, 2010).
For the past 10 years, blueberry producers in Michigan have increasingly adopted
integrated pest management (IPM) strategies, including reductions in the quantity and toxicity of
insecticides applied (Figure 2.1; NASS 2001-2011; Isaacs et al., 2009). The use of broadspectrum insecticides, such as azinphos-methyl, malathion, and carbaryl, declined from 20012011, in favor of reduced-risk insecticides such as acetamiprid and methoxyfenozide. However,
the recent arrival of spotted wing drosophila (Drosophila suzukii Matsumura), an invasive fruit
fly that can cause economic damage to stone and small fruits (Lee et al., 2011), threatens the
continued viability of IPM strategies in many fruit crops, particularly berry and cherry (Prunus
spp.) crops. In 2012, there were an estimated $207 million in losses from D. suzukii in the
eastern United States alone (eFly 2012). To prevent these losses, blueberry growers have

24

switched to repeated use of broad-spectrum insecticides applied on a calendar schedule,
generally every 7 days during the harvest period, to prevent infestation by D. suzukii (Van
Timmeren & Isaacs, 2013), a trend that may be highly deleterious to the bee species active in and
around crop fields after bloom.

Figure 2.1. Average percentage of bearing Michigan blueberry acres treated with each of the
eight conventional insecticides (solid line) and eight reduced-risk insecticides (dashed line)
reported by the National Agricultural Statistics Service (NASS 1991-2011). Conventional
insecticides included organophosphates, carbamates, and pyrethroids; reduced-risk insecticides
included biological insecticides, insect growth regulators, and neonicotinoids.

25

Community-level studies of wild bees have found that bees respond to resources and
stressors at multiple scales, with different species exhibiting differential responses to landscape
and other factors based on life-history traits such as body size, sociality, and nesting substrate
(Kleijn & van Langevelde, 2006; Brosi et al., 2007; Brittain et al., 2010a; Williams et al., 2010;
Watson et al., 2011; Sydenham et al., 2014). Crop flower visitation by bees declines with
increasing distance from natural and semi-natural habitats in a variety of agricultural systems
(Kremen et al., 2002; Hendrickx et al., 2007; Ricketts et al., 2008), but the scale at which bees
exhibit this response to habitat can differ based on individual species’ foraging ranges and diet
requirements (Cresswell et al., 2000; Steffan-Dewenter & Kuhn, 2003; Westphal et al., 2006).
Bee community responses to insecticide exposure may be similarly mediated by functional traits;
for example, smaller bees may have higher contact absorption of pesticides than larger bees due
to their higher surface area to volume ratio (Johansen, 1977; Brittain & Potts, 2011). Differences
in species-specific responses to insecticides based on functional or life-history traits could lead to
non-random changes in bee community composition in intensively managed landscapes (Brittain
& Potts, 2011). However, few studies have examined community-level responses of wild bees to
insecticides (but see Kevan et al., 1997; Brittain et al., 2010b; Tuell & Isaacs, 2010).
Understanding trait-based responses of bee communities to landscape composition and
insecticides will allow better prediction of the effects of land use change and agricultural
intensification on the pollination of crops and wild plants (Williams et al., 2010; Brittain &
Potts, 2011; Bartomeus et al., 2013).
Finally, in order to accurately estimate the effects of management practices and
conservation efforts at the local and landscape scales on pollination, it is necessary to identify the
main ‘ecosystem service providers’ and isolate their responses to these factors (Kremen, 2005;

26

Luck et al., 2009). Not all of the bee species captured by Tuell et al. (2009) in the Michigan
highbush agroecosystem provide pollination services to blueberry. Cleptoparasitic species, for
example, do not collect pollen because they are able to co-opt the pollen resources of their host
(Wcislo & Cane, 1996; Bogusch, 2003); while cleptoparasites have been found nectaring on
blueberry, they are unlikely to provide measurable pollination services. Isolating the responses of
service-providing bees to local and landscape factors will allow for better quantification of the
potential tradeoffs and cost/benefit analysis of different management practices and conservation
strategies for blueberry pollination (Luck et al., 2009).
In this study, I examined the major factors influencing wild bee abundance, richness, and
community composition in Michigan highbush blueberry fields during crop bloom in 3 years
prior to the arrival of D. suzukii (2004-2006) and 2 years following its arrival (2013-2014). My
objectives were (1) to determine whether insecticide use at the field scale and habitat
composition at the broader landscape scale are major drivers of wild bee abundance, richness,
diversity, and community composition in highbush blueberry fields, and (2) to determine
whether bee functional traits contribute to their responses to these factors. This study has broad
implications for the conservation of wild bee species for ecosystem service provision and
biodiversity in agricultural landscapes.

Materials and Methods
Study sites
The study was conducted in 15 highbush blueberry fields previously sampled by Tuell et
al. (2009), located in Allegan, Van Buren, and Ottawa counties in southwest Michigan (Figure
2.2). The lakeshore counties on the west coast of Michigan’s Lower Peninsula are characterized

27

by sandy, acidic soils and a lake-moderated microclimate that allows for a longer growing season
than other regions at the same latitude. This diverse and productive agricultural zone, sometimes
termed Michigan’s “Fruit Belt,” leads the nation in the production of highbush blueberries
(NASS 2011). The 15 sampled fields span a gradient of management intensity from unmanaged
fields to fields sprayed every 5-7 days for insect and disease control. Sites were located at least 3
km apart.

Figure 2.2. Map of the study area showing the location of highbush blueberry sites sampled in
2004-2006 (Tuell et al. 2009; Tuell & Isaacs 2010) and in 2013 and 2014 (this study).

28

Bee collections
To measure the abundance, richness, and community composition of wild bee
communities in the Michigan highbush blueberry agroecosystem, I used pan traps to collect bees
during blueberry bloom in 2013 and 2014 at the 15 sites previously sampled in 2004, 2005, and
2006 (reported in Tuell et al. 2009; Tuell & Isaacs 2010). In all years, bees were collected using
five pairs of pan traps (355 ml white and yellow plastic bowls; Amscan, Inc., Elmsford, NY)
placed 5 m apart along one transect at the field edge and a second transect 25 m into the field.
The twenty traps were half filled with a 2% soap solution (Dawn Ultra dish soap, Proctor &
Gamble, Cincinnati, OH) and mounted on 1.2 m PVC poles stabilized with rebar (Tuell & Isaacs,
2009). Traps were placed in the morning and collected after a minimum 6 h trapping period, 2-3
times during the bloom period depending on the duration of suitable weather conditions.
Specimens were placed in the freezer, then washed and dried as described in Tuell et al. (2009).
In addition, because pan trapping is likely to collect a biased subset of the total bee community
(Cane et al., 2001), wild bees were also hand netted from blueberry flowers for 10 minutes along
a transect from the field edge to 25m into the field interior at each site on the same day or the day
following pan trap collections.
Bee specimens (Hymenoptera: Apoidea) were identified to genus and species using
dichotomous keys (Mitchell, 1960, 1962; LaBerge, 1980; Michener et al., 1994; Gibbs, 2011;
Rehan & Sheffield, 2011) and the online keys available through www.discoverlife.org.
Specimens were compared with voucher specimens from Tuell et al. (2009) for verification.
Specimens from Tuell et al. (2009) in the Ceratina calcarata/dupla species complex were
reclassified according to Rehan and Sheffield (2011). Lasioglossum (subgenus Dialictus) and
Andrena species were identified by Dr. Jason Gibbs (Department of Entomology, Michigan State

29

University). Voucher specimens from this study were deposited in the Albert J. Cook Arthropod
Research Collection at Michigan State University.

Insecticide program risk
Growers provided spray records for the years 2003-2005 and 2012-2013 for 13 of 15
sites. Two sites provided records for four out of the five years. Prior-year insecticide applications
were used to calculate the per-site insecticide program risk (IPR) for each year, a metric used to
estimate the risk of insecticide use to bees described in Tuell and Isaacs (2010) (Equation 2.1).
Pesticide records from the year before bee collections were used because most insecticides are
applied after petal fall in blueberry, so effects of insecticides on the bee communities present
during bloom are likely to be seen the following season (Tuell & Isaacs, 2010). Fungicide
applications were not considered in this metric of insecticide risk.

(Equation 2.1) !!insecticide program risk! IPR !!=! Σ!

amount of active ingredient (kg)/Ha
contact LD50 for honeybees

Landscape characterization
Tuell (2007) digitized the landscape within a 1.5 km radius around the central pair of pan
traps in each field in ArcGIS (ArcMap 9.2, ESRI, Redlands, CA) using 2005 orthoimagery at the
1:2000 scale (National Agriculture Imagery Program (NAIP), accessed via the NRCS Geospatial
Data Gateway). These maps were updated using 2014 NAIP orthoimagery at the same scale in
ArcMap 10.1 (ESRI, Redlands, CA). Habitat types were classified into categories (Table 2.1)
and ground-truthed during site inspections in 2014 and using GoogleTM Street View. Concentric
circles with radii of 300, 500, 1000, and 1500 m were overlaid on the digitized landscape from
30

2005 and 2014 in order to calculate the proportion of habitat types surrounding each site at
different scales in the two years.

Table 2.1. Categories of habitat types used in landscape characterization (from Tuell, 2007).
Habitat type
Description
Blueberry plantations
commercial and semi-abandoned highbush blueberry
Perennial crops
other perennial crops, including vineyards and nurseries
Annual crops
field and vegetable crops
Pastures
grazing pastures
Open uncultivated
meadows, scrubland, fallow and other ruderal areas
Ditches and treelines
running along or bisecting agricultural land
Other field margins
margins along agricultural land other than ditches and tree lines
Forest/woodland
<10 m from forest edge
margin
Forest/woodland
>10 m from forest edge
interior
Settlement
suburban development including golf courses and a landfill
Road
paved or dirt
Railroad tracks
abandoned or still in use
Utility
areas cleared for powerlines
Shoreline
vegetation along Lake Michigan
Wetlands
vegetation on periodically flooded land
Riparian areas
vegetation along inland bodies of water such as ponds and lakes
Open water
open bodies of water including lakes, ponds, and river

Functional traits
A database of species functional traits for the collected specimens, including sociality,
nesting substrate, diet specialization, and intertegular distance, was compiled primarily from
Wolf & Ascher (2008) and the supplementary materials in Bartomeus et al. (2013), with
additional floral records to determine diet breadth for several Lasioglossum species from Gibbs
(2011) (Appendix B). These ‘response traits’ were selected for their likely association with bees’
responses to environmental factors such as resource distributions (Lavorel & Garnier, 2002).
Body size (as measured by intertegular distance) and sociality affect how far bees can travel in
31

search of food resources (Gathmann & Tscharntke, 2002), and thus determine at what scale
individuals interact with the surrounding landscape. Additionally, these two traits are likely to
affect how bees respond to pesticide exposure (Brittain et al., 2011). Nesting substrate and diet
specialization describe two resource requirements that also affect how bees interact with
landscape composition, as certain nesting substrates or floral hosts may be limited in different
environments (Roulston & Goodell, 2011). Bees that nest in wood, for example, may have
difficulty locating nest sites in highly simplified landscapes with little forest cover (Williams et
al., 2010). While some of these traits may also affect whether and how much different bee
species contribute to crop pollination – for example, larger-bodied bees tend to be more efficient
pollinators (Larsen et al., 2005) – I was not explicitly considering the effect that these functional
traits might have on pollination services.
To examine how the diversity of functional traits in bee communities might be affected
by landscape composition and insecticide risk, I calculated an index of functional dispersion for
the collected bee communities in each site in each year using four response traits: sociality,
nesting location, diet specialization, and body size (Table 2.2; Laliberté et al., 2010). Functional
dispersion is calculated from the mean distance in multidimensional trait space of individual
species to the centroid of all species, weighted by relative species abundances (function dbFD in
R package FD; Laliberté et al., 2010). I used the cailliez correction for the non-Euclidean
distances in the trait distance matrix generated by inclusion of categorical traits (Cailliez, 1983).
To quantify the impact of individual traits on the response of functional dispersion to predictor
variables, I recalculated trait diversity with each of the traits excluded in turn and ran the
analyses with each iteration.

32

Table 2.2. Traits used to calculate dispersion of functional traits of bees in the samples collected
in blueberry fields. Body size was condensed into a categorical variable, using inter-tegular (IT)
distance where available (Bartomeus et al., 2013).
Trait
Categories
Sociality
Solitary
Eusocial
Facultatively eusocial
Cleptoparasite
Body size
Small (IT <1.5 mm)
Medium (IT 1.5-2.5 mm)
Large (IT >2.5 mm)
Lecty (diet breadth)
Oligolectic
Polylectic
Cleptoparasite
Nest location
Soil (excavate)
Soil (rent)
Pithy stem
Cavity
Wood
Cleptoparasite

Service providing bees
Not all of the species collected in pan traps are likely to provide pollination services to
highbush blueberry. To isolate the response of service-providing bees to the studied factors, I
compiled a list of the subset of collected species that have been recorded foraging on Vaccinium
corymbosum L. and close relatives (lowbush blueberry, Vaccinium angustifolium Aiton and
deerberry, Vaccinium stamineum L.) in the eastern United States and Canada (Cane et al., 1985;
Stubbs et al., 1992; MacKenzie & Eickwort, 1996; Stubbs & Drummond, 1997; Javorek et al.,
2002; Sheffield et al., 2003; Tuell et al., 2009; Adamson, 2011; Benjamin et al., 2014; MoisanDeserres et al., 2014). I excluded cleptoparasitic bees from this list; although some species of
Nomada and Sphecodes have been recorded nectaring on Vaccinium spp., these species do not
actively collect pollen and thus are unlikely to provide effective pollination. Because only a few
of the blueberry pollination studies identified all observed Lasioglossum (Dialictus) and Andrena
33

to species, this list of 80 out of the 150+ species collected may be a conservative estimate of the
total number of species that can provide pollination services to Vaccinium in Michigan
(Appendix B).

Data analysis
All data analysis was performed in R version 3.0.2 (R Core Team, 2013). Species
accumulation curves were used to determine whether the sampling effort was sufficient to
estimate the community of bees likely to be captured in pan traps and nets in this habitat (Gotelli
& Colwell, 2001; Tuell et al., 2009). Netting and pan trapped collections were pooled, and all
observations were then averaged by site by year to avoid pseudoreplication and to correct for
uneven numbers of sampling rounds in different years. Average bee species richness was
characterized for each site in each year using the Chao1 richness estimator, which estimates
unseen species based on the rarity of sampled species (function estimateR in R package vegan;
Chao, 1987; Oksanen et al., 2013). Bias-corrected Shannon’s diversity indices using the ChaoShen estimator for unseen species were also calculated for each site in each year (function
entropy.ChaoShen in R package entropy; Chao & Shen, 2003).
Spatial autocorrelation was assessed using Mantel tests to compare pairwise bee
community similarity indices (Morisita-Horn index) with pairwise geographic distances between
each of the fifteen blueberry farms in each year (function mantel in R package vegan, 1000
permutations). No spatial autocorrelation among sites was detected in any year (all p > 0.29).
Non-metric multidimensional scaling (NMDS) ordination was used to visualize
differences in species composition of the wild bee communities among sites and across years
(function metaMDS in R package vegan; Oksanen et al., 2013). Species abundance data were

34

square root transformed and then submitted to Wisconsin double standardization prior to
calculating Bray-Curtis dissimilarities in order to improve the quality of ordination (functions
wisc and vegdist in R package vegan; Oksanen, 2006). Because of the large number of zeroes in
the site-by-species matrix, Bray-Curtis dissimilarities were zero-adjusted using a dummy species
with a value of 1 for all sites (Clarke et al., 2006). NMDS ordination was used to show the
trajectory of bee communities at individual sites over time (function ordiarrows in R package
vegan; Figure 2.6). I also fitted vectors of species loadings on the NMDS axes for the most
abundant species across all five years to examine possible trends related to functional traits
(function envfit in R package vegan), as well as 95% confidence ellipses around the group
centroids for different groups of species functional traits (function ordiellipse in R package
vegan).
Insecticide program risk (IPR) was not normally distributed, so I compared IPR across
years using a non-parametric Kruskal-Wallis test with post-hoc Dunn’s test with Bonferroni
correction. Two sites did not provide spray records in one of the five years; however, the Dunn’s
test for the median difference between groups is also appropriate for unbalanced designs.
Permutational analysis of variance (perMANOVA), a nonparametric equivalent of
multivariate analysis of variance that generates p-values using permutations and therefore does
not require the data to meet assumptions of normality, was used to compare bee community
composition across years (function adonis in R package vegan). Multivariate homogeneity of
variances was confirmed prior to perMANOVA analysis (function betadisper in R package
vegan). Similarity percentage (SIMPER) analysis was used to determine which species made the
largest contributions to the contrasts between years (function simper in R package vegan).

35

Principal components analysis (PCA), an ordination technique that creates orthogonal
vectors to explain major axes of variation in multivariate data sets, was used to reduce the
dimensionality of the landscape variables in 2005 and 2014 (function prcomp in R package stats;
R Core Team, 2013). Typically only a small number of principal components are needed to
explain the majority of the variation in complex data sets (Gotelli & Ellison, 2004). Because
these axes are orthogonal to each other – and thus wholly uncorrelated – the use of principal
components instead of raw landscape variables avoids potential problems with multicollinearity
in regression analyses.
I conducted separate principal components analyses (PCAs) for each of the four
landscape radii (300, 500, 1000, and 1500 m) for the 2005 and 2014 landscape data. I then
examined each PCA individually to determine a) how many landscape components to extract for
analysis and b) what major axes of variation were explained by the extracted components. I used
a combination of visual examination of scree plots (function fviz_screeplot in R package
factoextra; Kassambara, 2015) and the broken stick model (function evplot; Borcard et al., 2011)
to determine, for each model, which components retained valuable landscape information
(Jackson, 1993). These methods indicated consistent support for extracting at least the first two
principal components for all PCA models, with some support for extraction of one or two
additional axes depending on the landscape radius and year.
Bee species respond to habitat at different spatial scales based on life-history traits such
as diet breadth and body size, which determine resource requirements as well as maximum flight
radius and therefore which resources they can access in the landscape (Kreyer et al., 2004;
Greenleaf et al., 2007; Osborne et al., 2008; Jha & Kremen, 2013a; Wright et al., 2015). To
identify the scale at which surrounding landscape composition had the most explanatory power, I

36

used linear regressions of average bee abundance and estimated species richness against the first
two principal components for landscape variables at radii of 300, 500, 1000, and 1500 m, and
compared the resulting R2 values (Table 2.3; Steffan-Dewenter et al., 2002; Holland et al., 2004).
These regressions were conducted separately for abundance and richness pooled from 2004-2006
samples and 2013-2014 samples on the landscape data from 2005 and 2014, respectively. R2
values peaked at the 300 m radius for abundance in 2005 and 2014 and for richness in 2005
(Table 2.3), so only the 300 m scale landscape components were retained for use in subsequent
analyses.

Table 2.3. Linear regressions of pooled average bee abundance and estimated richness from
2004-2006 and 2013-2014 on the first two principal components from 2005 and 2014 landscape
data at 300m, 500m, 1000m, and 1500m radii around sampled sites.
Abundance
Richness
Landscape
2
2
radius
R
F
p
R
F
p
2005
300m
0.29
2.48
0.13
0.44
4.62
0.03*
500m
0.20
1.53
0.26
0.11
0.73
0.50
1000m
0.07
0.46
0.64
0.01
0.05
0.95
1500m
0.04
0.26
0.78
0.03
0.17
0.84
2014
300m
0.37
3.46
0.06
0.02
0.14
0.87
500m
0.28
2.33
0.14
0.07
0.44
0.66
1000m
0.23
1.76
0.21
0.05
0.32
0.74
1500m
0.18
1.34
0.3
0.02
0.12
0.88
* significant at p = 0.05

37

Examination of the biplots and factor loadings for the first two principal components for
landscape composition at the 300 m scale indicated that the major axes of variation explained by
these components were the same for 2005 and 2014 (Figures 2.9 and 2.10; Table 2.4). Linear
regressions of the first two PCA components from the 2005 and 2014 landscape analyses showed
strong positive correlations for both PC1 and PC2 (both R2 > 0.88 and p < 0.0001). The first two
components from the 2005 and 2014 landscape PCAs were therefore combined for repeated
measures analysis, with the 2005 component values assigned to sites for the years 2004-2006 and
the 2014 component values assigned to sites in 2013-2014.
To examine variability in responses to insecticide program risk by year, I ran simple
linear regressions of wild bee abundance and Chao1 estimated species richness on prior-year IPR
for each year of bee collections (2004, 2005, 2006, 2013, and 2014). Bee abundance was logtransformed prior to analysis to meet the assumption of normality. Only 14 sites were included in
the 2013 and 2014 analyses due to missing spray records from one site in each of those years. To
determine if bee responses to insecticide risk differed by collection period as well as over all five
years of collections, I performed separate repeated measures analyses of bee abundance and
richness on prior-year IPR for the 2004-2006 and 2013-2014 collections, as well as for all five
years of collections (function lme in R package nlme; Laird & Ware, 1982; Pinheiro et al., 2014).
Because of the non-continuous time series for bee collections (e.g. the five-year gap between the
2006 and 2013 collections), the repeated measures analyses were conducted with a variancecovariance matrix using a linear spatial correlation structure to account for the distance between
collection years (function corLin in R package nlme: Pinheiro et al., 2014). In addition, variances
were modeled explicitly for each year to account for heterogeneous variances for the response

38

variables in different years (function varIdent in R package nlme; Galecki & Burzykowski,
2013).
Following these analyses of IPR in different years and collection periods, repeated
measures linear mixed models were used to determine the combined effect of IPR and landscape
factors on the abundance, richness, diversity, and functional dispersion of bees across all five
years of collections (function lme in R package nlme; Laird & Ware, 1982; Pinheiro et al., 2014).
The two sites with only four years of insecticide records were removed prior to repeated
measures analyses, which require a balanced variance-covariance matrix. These analyses were
conducted with an unstructured variance-covariance matrix with heterogeneous variances by
year. Variance components were estimated using maximum likelihood. Bee abundance data were
log-transformed prior to analysis to meet the assumption of normality.
Automated model selection for each response variable was performed using the dredge
function in R package MuMIn, which calculates Akaike’s information criterion (AIC) values for
models using all possible combinations of predictor variables and ranks them based on AIC
(Barton, 2014). The relative importance of each predictor variable was then calculated from the
sum of AIC weights for all models with delta AIC (ΔAIC) ≤ 2 containing the variable (function
importance in R package MuMIn). Delta AIC is a measure of each model relative to the model
with the lowest AIC value (the ‘best’ model in terms of goodness of model fit with a minimum
number of model parameters); according to a rule of thumb provided by Burnham & Anderson
(2002), there remains substantial evidence for models within 2 AIC values of the ‘best’ model,
while ΔAIC > 10 indicates that the model is very unlikely.
I ran the five-year repeated measures analyses again with only the subset of wild bees
that could potentially provide pollination services in blueberry (Appendix B). I also compared

39

the responses of social bee and solitary bee abundance and estimated richness in separate
regression models to determine whether sociality affects the response of bee species to the
predictor variables. Abundances of social and solitary species were log(x + 1) transformed prior
to analysis to meet the assumption of normality. Following this analysis, I ran separate analyses
for the 2004-2006 and 2013-2014 collection periods to rule out the possibility that the patterns
were related to extreme weather between the two periods.
Following mixed model analyses, I fitted vectors of significant landscape variables and
insecticide program risk from the mixed models onto the NMDS ordination of bee communities
among sites using a multivariate correlation analysis that partitions the linear component of each
predictor on the NMDS axes (function envfit in R package vegan; Williams & Winfree, 2013).
When ordination stress is low, this approach can be used to test the effect of predictor variables
on community composition.
Results
Bee collections
In 2013, I collected 793 bees during three bloom samples at each of the 15 sites,
representing at least 81 species, 17 genera, and 5 families. These samples included 372 honey
bees (Apis mellifera L.) and 421 wild bees (Appendix B). In 2014, I collected 1,829 bees
representing at least 90 species, 17 genera, and 5 families, including 1,055 A. mellifera. Tuell et
al. (2009) collected a total of 3,228 non-Apis bees at these sites from 2004-2006 over seven
bloom samples (2-3 samples per year per site), representing at least 112 species, 20 genera and 5
families (Tuell et al. 2009). The species accumulation curve calculated across all five years of
sampling, with one sample round per site (20 pan traps and 10 minutes of net collecting) as the
base sampling unit, indicated that the number of species collected begins to level off after the
40

equivalent of one year’s worth of sampling at 15 sites (n = 45 samples) (Figure 2.3). Species
accumulation curves for individual years showed differences in total asymptotic richness
collected in different years, with lower asymptotic richness in 2013 and 2014 than in the 20042006 samples (Figure 2.4).

Figure 2.3. Species accumulation curve generated from bee community samples during bloom in
2004-2006 and 2013-2014 at 15 blueberry farms. One sample represents the community of bees
collected in one day at one site (20 pan traps and 10 minutes of hand netting). Dashed line
represents the 95% confidence interval around the mean.

41

Figure 2.4. Species accumulation curves for each year of sampling. One sample represents the
community of bees collected in one day at one of the 15 sites (20 pan traps and 10 minutes of
hand netting). At each site, there were two sampling rounds during bloom in 2004 and 2006 (30
samples) and three sampling rounds in 2005, 2013, and 2014 (45 samples). Error bars represent
95% confidence around the mean line for each additional sample.

The most abundant bee captured in pan traps in all years (2004-06, 2013-14) was A.
mellifera, the managed European honey bee. While the Vaccinium specialist Andrena carolina
Viereck represented 16% of all wild bee abundance in the 2004-2006 collections by Tuell et al.
(2009), only 8 specimens of A. carolina were captured in 2013 and 16 specimens in 2014,
representing 1.9% and 2.1% of wild bee abundance, respectively. The most abundant non-Apis
bees in 2013 and 2014 were several social or semisocial halictid and apid species, including
Lasioglossum pilosum (Smith), Lasioglossum coriaceum (Smith), Bombus impatiens Cresson,
Augochlorella aurata (Smith), Lasioglossum cressonii (Robertson), and Lasioglossum imitatum
(Smith).
42

Nonmetric multidimensional scaling (NMDS) ordination of Bray-Curtis dissimilarities
between bee communities at each site in each year suggested that the communities in the first
three years of sampling were more similar to each other than to the communities collected in
2013 and 2014 (k =2, stress = 0.25; Figure 2.5). Total community composition across sites
differed significantly among years (F = 3.32, df = 4, 70, p = 0.001). Individual sites had high
species turnover from year to year, resulting in dissimilar communities from year to year in
ordination space (Figure 2.6).

Figure 2.5. Nonmetric multidimensional scaling ordination of bee communities collected in pan
traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014. Each point represents one
site in one year, with different colors for each of the five years. Ellipses represent 95%
confidence ellipses around the group mean for each year.

43

!!!!
Figure 2.6. Nonmetric multidimensional scaling ordination of bee communities collected in pan
traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014. Each point represents one
site in one year. Black arrows show the trajectory of one site beginning in 2004 (circled points)
and ending in 2014.
Insecticide program risk
The 15 farms varied in both the types and rates of insecticides used in all years, resulting
in a wide range of insecticide program risk (IPR) scores within and across years (Figure 2.7).
The median IPR score differed across years (χ2 = 12.8, df = 4, p = 0.01), increasing from 2005 to
2012 with additional increase in 2013. Most growers did not apply insecticides during the bloom
period, but began spraying more intensively during the berry sizing and pre-harvest periods to
control for blueberry maggot (Diptera: Tephritidae; Rhagoletis mendax Curran), Japanese beetle
(Coleoptera: Scarabaeidae; Popillia japonica Newman), and D. suzukii, according to grower
records (Figure 2.8).
44

Figure 2.7. Boxplot of insecticide program risk across 15 blueberry farms in each of the five
years prior to bee collections.

Figure 2.8. Increase in cumulative insecticide program risk (IPR) value over the 2012 growing
season at the 13 conventionally managed blueberry fields. The average cumulative IPR is
represented by the black dashed line.
45

Landscape characterization
Sites varied in the proportion of each land-use type in the surrounding landscape, but
experienced relatively little land use change from 2005 to 2014. Blueberry plantings were the
dominant land cover category within a 300 m radius of the sampled fields, comprising
approximately 29-30% of the area within 300 m. This proportion stayed constant from 2005 to
2014 when averaged across sites. Individual sites had larger changes at this scale; two farms
cleared several acres of forest to plant additional blueberry, while another farm removed older
stands of blueberry and replaced them with row crops. Forest interiors, which represented around
24% of land cover in 2005, decreased by only about 1% over all sites from 2005-2014, with
around 16 acres of forest cleared for agricultural and development purposes. However, forest
margins (the first 10 m edge of forest patches) increased from 6 to 8% of total land area as these
larger forest patches were cleared. Settled areas, which represented around 10% of land cover in
2005, increased by 14 acres, or around 2%. Row crops represented 4% of land cover at the 300
m scale in 2005, increasing by 11.5 acres to 5% of land cover in 2014.
The first two principal components in the PCA of landscape at the 300 m scale around the
sample sites represented tradeoffs in four of the dominant categories of land cover directly
surrounding sampled fields: blueberry plantings, forest patches, row crops, and settled areas
(Figures 2.9 and 2.10; Table 2.4). The first principal component for landscape at 300 m was
strongly associated with the proportion of blueberry plantings in the surrounding landscape
(Table 2.4). Several of the sampled fields were part of large commercial blueberry farms that
included many acres of blueberry within a 300 m radius. In addition, the area of forest interior
(more than 10 m into forest patches) and row crops, which consisted primarily of corn and
soybean fields, loaded negatively on this first component. This axis (PC1) represents the main

46

axis of variation in the landscape directly surrounding the sampled fields: as the amount of
blueberry in the landscape increases, large forest patches and other agricultural crops decrease.
The second principal component for landscape at 300 m was also strongly negatively
associated with forest interior (Table 2.4); however, unlike the first principal component,
blueberry plantings also loaded negatively on this axis. Settled areas – comprised primarily of
houses, farm buildings, mowed lawns, and driveways – had a strong positive loading on this
component. PC2 therefore represents the second dominant axis of variation in the landscape
directly surrounding blueberry fields: as the amount of developed land increased, the area of
forested patches and blueberry plantings decreased.

47

Figure 2.9. Biplot of first two principal components for landscapes surrounding blueberry field
locations sampled for bees at the 300 m scale in 2005.

Figure 2.10. Biplot of first two principal components for landscapes surrounding blueberry field
locations sampled for bees at the 300 m scale in 2014.
48

Table 2.4. Factor loadings for the first two principal components for landscape in 2005 and
2014. Loadings > 0.4 are shown in bold.
2005
2014
Land use
PC1
PC2
PC1
PC2
0.83
-0.30
0.81
-0.25
Blueberry
-0.13
0.08
-0.04
0.05
Other perennial crops
-0.09
0.05
0.00
0.01
Pasture
-0.27
0.23
-0.36
0.26
Row crops
-0.18
-0.05
0.04
0.09
Open uncultivated
0.06
0.00
0.06
0.01
Ditches
-0.18
0.11
-0.19
0.14
Field margins
-0.33
-0.77
-0.40
-0.80
Forest interior
-0.01
0.05
-0.02
-0.01
Forest margins
0.02
0.05
0.01
0.04
Road
0.17
0.48
-0.02
0.43
Settled
0.02
-0.01
0.02
0.00
Railroad track
0.00
0.00
0.00
0.00
Riparian areas
0.11
0.07
0.10
0.05
Water

Bee responses to pest management and landscape
Simple linear regressions of log-transformed wild bee abundance on IPR found
significant negative relationships between abundance and prior-year IPR for the 2004 and 2014
collections (Table 2.5). Relationships in the other three years were more variable, including a
positive trend between IPR and abundance in 2006 (Figure 2.11). Linear regressions of Chao1
estimated species richness of wild bees on prior-year IPR for each year of collections found a
marginally significant negative correlation in 2004 and a significant negative correlation in 2013
(Table 2.5). The relationship trended negative in 2005 and 2014 and – like abundance – showed
a positive trend in 2006 though this had a weak correlation (Figure 2.12).
When abundance and richness were analyzed against prior-year IPR separately for the
two collection periods (2004-2006 and 2013-2014) and then for all five years of sampling, only
the five year repeated measures analyses found significant negative relationships with IPR (Table
49

2.6). The slope of the relationship between bee abundance and IPR was positive for the 20042006 collection period and negative for the 2013-2014 collections (Table 2.6), though neither
were significant. Wild bee richness had a marginally significant negative correlation with prioryear IPR for the 2004-2006 period (F = 3.57, df = 1,29, p = 0.07; Table 2.6), but was less
correlated with IPR in 2013-2014 (F = 2.48, df = 1,12, p = 0.14). However, this relationship was
highly significant when analyzed across all five years (F = 13.6, df = 1,51, p = 0.0005).
Log-transformed bee abundance across all five years was significantly negatively
correlated with IPR and the second principal component for landscape at 300 m, which was
associated with increasing settlement and decreasing forest cover (Tables 2.7 and 2.8; Figures
2.13 and 2.14). Bee richness (Chao1) was also significantly negatively correlated with IPR
(Tables 2.9 and 2.10; Figure 2.15). The competing models for bee richness included one or both
landscape components, both of which were negatively correlated with bee richness (Table 2.9).
Bias-corrected Shannon’s diversity of wild bees had a significant negative relationship
with insecticide program risk, though the null model also had substantial support according to
ΔAIC (Tables 2.11 and 2.12; Figure 2.16). However, functional dispersion (FDis) was not
explained well by any of the predictor variables (Table 2.13). Akaike weights indicated little
confidence in any model of functional dispersion (all models <22% confidence; Table 2.13). No
combination of functional traits (sociality, body size, lecty, or nest location) substantially
improved the model fit for FDis; the null model was either the best model or within 0.34 ΔAIC
for all combinations of FDis.

50

Table 2.5. Slopes (β ± SE), coefficients of determination (R2), and p-values for simple linear
regressions of log-transformed wild bee abundance and Chao1 estimated species richness on
prior-year insecticide program risk for each year of collections at 13 blueberry farms.
Log-transformed abundance
Chao1 estimated species richness
Collection
2
year
β ± SE
R
p
β ± SE
R2
p
2004
-0.18 ± 0.05
0.5
0.003**
-0.81 ± 0.43
0.21
0.09
2005
-0.08 ± 0.05
0.16
0.15
-0.69 ± 0.43
0.17
0.13
2006
0.05 ± 0.04
0.09
0.28
0.36 ± 0.57
0.03
0.54
2013
0.001 ± 0.02
0.0001
0.97
-0.34 ± 0.12
0.41
0.01*
2014
-0.04 ± 0.02
0.33
0.03*
-0.33 ± 0.26
0.12
0.23
*significant at p=0.05; **p=0.01

Table 2.6. Slopes (β ± SE), degrees of freedom (df), and p-values for repeated measures
regressions of log-transformed wild bee abundance and Chao1 estimated species richness on
prior-year insecticide program risk for 2004-2006 collections, 2013-2014 collections, and all five
years of collections at 13 blueberry farms.
Log-transformed abundance
Chao1 estimated species richness
Collection
period
β ± SE
df
p
β ± SE
df
p
2004-2006
0.02 ± 0.02
29
0.41
-0.58 ± 0.31
29
0.07
2013-2014
-0.02 ± 0.02
12
0.22
-0.26 ± 0.17
12
0.14
All years
-0.03 ± 0.01
51
0.046*
-0.46 ± 0.12
51
0.0005***
*significant at p=0.05; **p=0.01; ***p=0.001

51

!
Figure 2.11. Simple linear regressions of log-transformed abundance of wild bees on prior-year insecticide program risk (IPR) at the
13 blueberry farms that provided 5 years of application records. Dashed lines represent regression trend lines.

Figure 2.12. Simple linear regressions of Chao1 estimated species richness of wild bees on prior-year insecticide program risk (IPR)
at the 13 blueberry farms that provided 5 years of application records. Dashed lines represent regression trend lines.
52

Table 2.7. Model selection table for log-transformed wild bee abundance based on repeated
measures analysis on all combinations of insecticide program risk (IPR) and the two first
principal components for landscape composition (300 m radius). Models are ranked according to
Akaike’s information criterion (AIC), with the regression slopes, AIC, ΔAIC, AIC weight, and
relative variable importance based on Akaike weights shown for the best competing models
(ΔAIC ≤ 2) only.
Model
rank

Model components
IPR

PC1

1
-0.03*
n.i.
2
n.i.
n.i.
3
-0.03* -3.8x10-6
Importance 0.70
0.25
n.i. = not included in model.
*significant at p = 0.05.

k

PC2
-5.0x10-6*
-5.7x10-6*
-5.5x10-6*
1.00

2
1
3

Akaike’s
information
criterion (AIC)
154.7
155.5
155.9

ΔAIC

AIC weight

0
0.84
1.43

0.455
0.299
0.198

Table 2.8. Best model for log-transformed wild bee abundance across five years of sampling
(2004-2006 and 2013-2014) at 13 blueberry farms.
Parameter
IPR
PC2

Model coefficient ± SEM
-0.03 ± 0.01
-0.00001 ± 0.00

df
50
50

t-value
-2.02
-2.29

p-value
0.048*
0.03*

*significant at p = 0.05.
Table 2.9. Model selection table for Chao1 estimated richness for wild bees based on repeated
measures analysis on all combinations of insecticide program risk (IPR) and the two first
principal components for landscape composition (300 m radius). Models are ranked according to
Akaike’s information criterion (AIC), with the regression slopes for model components, AIC,
ΔAIC, AIC weight, and relative variable importance based on Akaike weights shown for the best
competing models (ΔAIC ≤ 2) only.
Akaike’s
Model components
Model
k
information
ΔAIC AIC weight
rank
IPR
PC1
PC2
criterion (AIC)
1
-0.46***
n.i.
n.i.
1
470.6
0
0.451
-5
2
-0.48*** -3.5x10
n.i.
2
471.4
0.78
0.305
-5
3
-0.43**
n.i.
-1.7x10
2
471.8
1.22
0.245
Importance
1.00
0.30
0.24
n.i. = not included in model.
*significant at p = 0.05; **p = 0.01; *** p = 0.001.

53

Table 2.10. Best model for Chao1 estimated wild bee richness across five years of sampling
(2004-2006 and 2013-2014) at 13 blueberry farms.
Parameter
Model coefficient ± SEM
df
t-value
p-value
IPR
-0.46 ± 0.12
51
-3.69
0.0005***
***significant at p = 0.001.

Table 2.11. Model selection table for bias-corrected Shannon’s diversity of wild bees based on
repeated measures analysis on all combinations of insecticide program risk (IPR) and the two
first principal components for landscape composition (300 m radius). Models are ranked
according to Akaike’s information criterion (AIC), with the regression slopes for model
components, AIC, ΔAIC, AIC weight, and relative variable importance based on Akaike weights
shown for the best competing models (ΔAIC ≤ 2) only.
Akaike’s
Model components
Model
k
information
ΔAIC AIC weight
rank
IPR
PC1
PC2
criterion (AIC)
1
-0.01**
n.i.
n.i.
1
13.9
0
0.397
-7
2
-0.01*** 9.7x10
n.i.
2
14.7
0.82
0.263
3
n.i.
n.i.
n.i.
0
15.3
1.43
0.194
4
-0.01**
n.i.
2.6x10-8 2
15.9
2.00
0.146
Importance
0.81
0.26
0.15
n.i. = not included in model.
*significant at p = 0.05; **p = 0.01; *** p = 0.001.

Table 2.12. Best model for bias-corrected Shannon’s diversity of wild bees collected across five
years of sampling (2004-2006 and 2013-2014) at 13 blueberry farms.
Parameter
Model coefficient ± SEM
df
t-value
p-value
IPR
-0.009 ± 0.003
51
-3.05
0.004**
**significant at p = 0.01.

54

Table 2.13. Model selection table for functional dispersion (FDis) of wild bees based on
repeated measures analysis on all combinations of insecticide program risk (IPR) and the two
principal components for landscape composition (300 m radius). Models are ranked according to
Akaike’s information criterion (AIC), with the regression slopes for model components, AIC,
ΔAIC, AIC weight, and relative variable importance based on Akaike weights shown for the best
competing models (ΔAIC ≤ 2) only.
Akaike’s
Model components
Model
k
information
ΔAIC AIC weight
rank
IPR
PC1
PC2
criterion (AIC)
1
n.i.
-4.7x10-7
n.i.
1
-171.7
0
0.216
2
n.i.
n.i.
n.i.
0
-171.3
0.31
0.185
-7
3
0.001 -4.4x10
n.i.
2
-171.3
0.36
0.180
4
0.002
n.i.
n.i.
1
-171.1
0.52
0.167
-7
-8
5
n.i.
-4.6x10 -7.8x10
2
-169.8
1.81
0.087
6
0.002 -4.3x10-7 -1.2x10-7
3
-169.8
1.90
0.083
-7
7
0.002
-1.4x10
1
-169.7
1.96
0.081
Importance
0.51
0.57
0.25
n.i. = not included in model.

Figure 2.13. Average untransformed abundance of wild bees per site captured over five years of
sampling during bloom at 13 blueberry farms, plotted against the average prior-year insecticide
program risk at those farms. Gray area represents 95% confidence around the regression line.
Two farms that did not provide all five years of spray records were not included in the regression
analysis and are not shown here.
55

Figure 2.14. Average untransformed abundance of wild bees per site captured over five years of
sampling during bloom at 13 blueberry farms, plotted against the second principal component for
landscape at 300 m (PC2) at those farms. Increasing PC2 is associated with decreasing forest
cover and increasing area of settlement (farm buildings, houses, lawns, driveways, etc.). Grey
area represents 95% confidence around the regression line. Two farms that did not provide all
five years of spray records were not included in the regression analysis and are not shown here.

Figure 2.15. Repeated measures regression of prior-year insecticide program risk on Chao1
estimated species richness of the wild bees captured over 5 years of sampling during bloom at 13
blueberry farms. Grey area represents 95% confidence around the regression line. Two farms
that did not provide all five years of spray records were not included in the regression analysis
and are not shown here.
56

Figure 2.16. Average bias-corrected Shannon’s diversity of the wild bees captured over five
years of sampling during bloom at 13 blueberry farms, plotted against the average prior-year
insecticide program risk at those farms. Two farms that did not provide all five years of spray
records were not included in the regression analysis and are not shown here.

When vectors of insecticide program risk and the two landscape components were fitted
onto the NMDS ordination of bee communities, both IPR and PC1, or increasing area of
blueberry in the landscape, showed significant correlation with community composition (Figure
2.17; Table 2.14). The second principal component for landscape, which represented increasing
settlement and decreasing forest cover, was not significantly correlated with community
composition (Table 2.14).

57

!
Figure 2.17. Nonmetric multidimensional scaling ordination of bee communities collected in
pan traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014. Each point represents
one site in one year. Black arrows show the trajectory of one site beginning in 2004 (circled
points) and ending in 2014. Red arrows show the projected vectors of insecticide program risk
(IPR) and the two first principal components for landscape at the 300 m scale. Arrow lengths for
IPR, PC1, and PC2 are scaled based on correlation with the ordination configuration.

Table 2.14. Direction cosines and correlations of the three predictor variables with the twodimensional NMDS ordination of bee community composition.
Direction cosines
Predictor
r2
p-value
NMDS1
NMDS2
variable
IPR
0.77
-0.64
0.20
0.002
PC1
0.97
0.26
0.02
0.45
PC2
0.55
0.84
0.26
0.001

58

Comparison of social and solitary species
Visual examination of the NMDS ordination with fitted vectors of species loadings for
the most abundant wild bee species collected across all years suggested that the most abundant
solitary species (Andrena spp. and Lasioglossum leucozonium (Schrank)) were more associated
with the first three years of collection, while the most abundant social, subsocial, and
facultatively social species were more associated with the 2013 and 2014 collections
(Lasioglossum (Dialictus) spp., B. impatiens, Ceratina calcarata Robertson, Augochlora pura
(Say), and Augochlorella aurata (Smith)) (Figure 2.18). Across all collected species, solitary and
social species grouped separately in ordination space (Figure 2.19).

!
Figure 2.18. Biplot of NMDS ordination of bee communities collected from 2004-2006 and
2013-2014, showing the projection of fitted vectors for the most abundant species across all five
years. Each colored dot represents one site in one year. Andrena imitatrix and Lasioglossum
cressonii, which overlapped strongly with A. miserabilis and L. leucococum, respectively, are not
shown.
59

!!
Figure 2.19. Nonmetric multidimensional scaling ordination of bee communities collected in
pan traps and nets in 15 blueberry fields from 2004-2006 and 2013-2014. Each colored dot
represents one site in one year. Open black circles represent a single species across all five years.
Ellipses represent 95% confidence ellipses around the group mean for solitary and social
(including social, subsocial, and facultatively social) bee species.
Social (including eusocial, facultatively social, and subsocial) and solitary bee species
varied in their responses to IPR and landscape variables. Log-transformed abundance of solitary
species had a significant negative correlation with PC1 (F = 7.41, df = 1,49, p = 0.009) and a
marginally significant negative relationship with IPR (F = 2.75, df = 1,49, p = 0.09). However,
while log abundance of social species was also negatively correlated with PC1 (F = 4.90, df =
1,49, p = 0.03), it showed no correlation with IPR (F= 0.13, df = 1,49, p = 0.72). Species
richness of solitary species was strongly negatively correlated with IPR (Figure 2.19; F = 12.43,
df = 1,49, p = 0.0009) and PC2 (F = 6.72, df =1,49, p = 0.01). Social bee species richness was

60

also significantly negatively correlated with IPR (Figure 2.19; F = 5.16, df = 1,49, p = 0.03), and
had a marginally significant negative correlation with PC1 (F= 3.01, df = 1,49, p = 0.08).
When analyzed separately, the 2004-2006 collections showed slightly different patterns
than the analyses across all years. Log abundance of solitary bees was significantly negatively
correlated with IPR (F = 10.92, df = 1,29, p = 0.003), and had a marginally significant negative
correlation with PC2 (F = 4.67, df = 1,12, p =0.052). Social bee abundance was marginally
negatively correlated with PC1 (F = 3.64, df = 1,12, p = 0.08), but not IPR or PC2. Solitary
richness was significantly negatively correlated with PC2 (F = 8.31, df = 1,12, p = 0.01), but was
not significantly correlated with IPR (F = 2.57, df = 1,29, p = 0.12) or PC1 (F = 3.09, df = 1,12,
p = 0.10). Social bee richness was negatively correlated with PC1 (F = 7.17, df = 1,12, p = 0.02)
and positively correlated with PC2 (F = 6.47, df = 1,12, p = 0.03), but was not correlated with
IPR (F = 1.99, df = 1,29, p = 0.17).
The 2013-2014 collections also showed different patterns than all five years together.
Log abundance of solitary bees had a significant negative correlation with PC1 (F = 5.40, df
=1,10, p = 0.04) and a marginally significant negative correlation with IPR (F = 4.56, df = 1,12,
p = 0.054). Similar to the 2004-2006 collection years, log abundance of social species had a
marginally significant negative correlation with PC1 (F = 4.82, df = 1,10, p = 0.053), but was not
correlated with IPR or PC2. Solitary richness was significantly negatively associated with IPR (F
= 6.43, df = 1,12, p = 0.03), but was not associated with either landscape variable. Species
richness of the 2013-2014 social species was not associated with any of the measured variables.

61

Figure 2.20. Repeated measures regressions of the average species richness of social (blue line)
and solitary (yellow line) wild bee species on the prior-year insecticide program risk at 13
blueberry farms over five years of collections.
Service providing bees
Overall community responses to pest management and landscape composition were
similar for the subset of 80 wild bee species that have been recorded foraging on blueberry. Logtransformed abundance for these bees was significantly negatively correlated with the first
principal component for landscape (F = 7.95, df = 1,49, p = 0.007), and had a marginally
significant negative correlation with IPR (F = 3.36, df = 1,49, p = 0.07). Chao1 estimated species
richness of service-providing bees was significantly negatively correlated with IPR (F = 5.89, df
= 1,49, p = 0.02), but not either landscape variable. Bias-corrected Shannon’s diversity for the
service-providing bees had a marginally significant negative correlation with PC2 (F = 2.94, df =
1,49, p = 0.09), but was not correlated with IPR or PC1.

62

Discussion
The agricultural matrix of the Michigan highbush blueberry agroecosystem consists
largely of intensively managed crop fields surrounded by other blueberry fields, strips of grass,
and patches of forest. Despite this relatively resource-poor environment for wild bees, blueberry
fields consistently receive between 10 and 50% of their pollination from the wild bees living in
this matrix (Isaacs & Kirk, 2010). This study shows that the diverse communities of wild bee
visiting blueberry fields during bloom, including the subset of species that provide pollination
services to blueberry, respond strongly to local-scale factors, including insecticide program risk
and local landscape composition. In particular, the intensity of insecticide use at the field scale
has a strong negative effect on the abundance, richness, and diversity of wild bee communities
foraging in blueberry fields during bloom. Landscape composition around blueberry fields is also
associated with environmental stress on wild bee communities; wild bee abundance declined
with decreasing area of forest and the associated increase in nonagricultural developed areas.
This work builds directly on that of Tuell (2007), Tuell et al. (2009), and Tuell & Isaacs
(2010), by providing a longer-term evaluation of the community-level and species-specific
responses of wild bees to landscape composition and insecticide risk. Tuell & Isaacs (2010)
found that bee responses to insecticide risk differed both at the community and species levels
from year to year. High annual variability in the abundance, richness, and community
composition of bee communities mean that the results of short-term ecological studies of these
communities may not be representative of long-term trends (Roubik, 2001; Williams et al.,
2001). Using repeated measures analyses of bee communities at these sites, I was able to
determine that wild bee communities were more consistent in their responses to insecticide risk

63

over the sampled period than to landscape variables, although there was less change in land
cover than in insecticide use in this time frame. I was also able to establish that the total wild bee
community and the subset of species that provide pollination services to blueberry show similar
responses to insecticide risk and landscape; therefore, conservation strategies that benefit the
bees that provide pollination services to blueberry are also likely to support the overall
biodiversity of bees in this landscape, and vice versa.
Wild bee communities responded most strongly to the major axes of landscape
composition variation at the smallest scale studied (300 m). Bee foraging ranges vary based on
body size, sociality, and resource distributions in the landscape, with larger social bees, such as
some Bombus spp., traveling a kilometer or more in search of floral resources (Walther-Hellwig
& Frankl, 2000; Gathmann & Tscharntke, 2002; Greenleaf et al., 2007; Jha & Kremen, 2013a).
However, many solitary species have more restricted foraging ranges (Gathmann & Tscharntke,
2002; Greenleaf et al., 2007; Zurbuchen et al., 2010a, 2010b). Zurbuchen et al. (2010b) found
that while some small Hylaeus could travel a maximum distance of approximately 1 kilometer in
search of food, half of the studied females did not forage at distances longer than 100-300m.
Long foraging distances may impose greater costs on solitary species, which must return to their
nests several times a day to provision offspring, while social species can spread the energetic
costs of foraging flights across a greater number of individuals (Zurbuchen et al., 2010a, 2010b).
Over half of the bee species in this study were small solitary species likely to nest and forage
close to the capture site, which may explain why the community as a whole responded most
strongly to the smallest landscape scale encompassing all species’ foraging ranges.
Wild bee communities showed clear negative responses to prior-year insecticide program
risk (IPR) at the field scale across the five years of collections. While it is widely known –

64

primarily from laboratory studies – that many chemicals used for agricultural pest management
are toxic to bees at individual and colony levels (Helson et al., 1994; Thompson, 2003; Morandin
et al., 2005; Desneux et al., 2007; Mullin et al., 2010; Brittain & Potts, 2011; Gill et al., 2012;
Biddinger et al., 2013; Arena & Sgolastra, 2014), few studies have examined wild bee
community responses to pesticides (but see Kevan, 1975; Kevan et al., 1997; Dormann et al.,
2007; Brittain et al., 2010; Tuell & Isaacs, 2010). This study found that wild bees are less
abundant, species-rich, and diverse on farms with more intensive insecticide use, providing
support for the assertion of Goulson et al. (2015) that pesticides are an important contributing
factor to the declines in bee populations observed in various agricultural systems.
Bee abundance also increased with increasing forest cover and declined with increasing
settlement at the 300 m scale, as represented by PC2. These results align well with Watson et al.
(2011), who found that the abundance and richness of wild bees in Wisconsin apple orchards
during bloom were positively correlated with the proportion of forest cover and negatively
correlated with the proportion of nonagricultural developed land in the surrounding landscape.
Early spring bees, such as the mining bees (Andrena spp.) that are active pollinators of blueberry
and apple, may be associated with forested areas, which provide floral resources in the form of
flowering trees and shrubs at this time of year (Giles & Ascher, 2006; Westwood, 2006; Grundel
et al., 2010). Many wild bee species collect pollen from flowering trees, including those
commonly found in forest understories, such as those in the genera Salix, Crataegus, Cornus,
Acer, and Prunus (Batra, 1985; Stubbs et al., 1992; Giles & Ascher, 2006). In a season-long
survey of bees in different habitats in northwest Indiana, Grundel et al. (2010) found wild bees
were relatively abundant in forests in early to mid May, before the canopy filled in, but were not
abundant in these habitats later in the season. Following blueberry bloom, therefore, wild bee

65

communities may exhibit different relationships with land cover that reflect seasonal changes in
resource distributions.
Increasing area of nonagricultural developed land, including houses, driveways, farm
buildings, and mowed lawns around those built spaces, may be correlated with reductions in the
availability of floral resources or nesting sites for wild bees (Westrich, 1996; Kearns & Inouye,
1997). Conversion of land from natural habitat to built spaces may also lead to fragmentation of
plant communities and create barriers to pollinator movement (Westrich, 1996; Kearns &
Inouye, 1997; Steffan-Dewenter & Tscharntke, 1999; Hernandez et al., 2009). Bates et al. (2011)
found that wild bee abundance and richness along an urban to rural gradient declined with the
percentage of built space in the surrounding landscape. Ground-nesting bees respond negatively
to an increasing proportion of impervious surfaces in the surrounding environment (Ahrné et al.,
2009; Jha & Kremen, 2013a). Jha & Kremen (2013a, 2013b) found that nest density and gene
flow of bumble bees (Bombus spp.), which nest in the soil, decrease with an increasing
proportion of impervious surface in the landscape. Reductions in forest cover in favor of
increasing development may therefore have non-random effects on bee communities based on
life-history traits such as nesting substrate.
Functional trait diversity, as measured by functional dispersion (Laliberté et al., 2010),
remained stable along gradients of landscape composition and insecticide risk in this study.
However, sociality appears to mediate wild bee responses to these environmental factors, with
solitary species showing slightly more negative responses to insecticide risk than social species.
Abundance of solitary species, but not social species, was negatively correlated with IPR for
both the 2004-2006 and 2013-2014 collections, as well as for the bees from all five years
analyzed together. Both social and solitary bee richness were significantly negatively correlated

66

with IPR when analyzed across all years of the study, with solitary richness showing a stronger
correlation and steeper negative slope than social richness. However, neither social nor solitary
bee richness for the 2004-2006 bees were correlated with IPR, though both had significant
correlations with at least one landscape variable during this time period. Solitary bee richness,
but not social bee richness, was significantly negatively correlated with IPR in the 2013-2014
collections. The differences in species richness responses between these time periods may
indicate that the major environmental filter for these communities has shifted from the resource
distribution in the landscape (2004-2006) to the distribution of environmental stress (2013-2014).
The observed pattern of greater effect of insecticide risk on solitary species abundance
and richness was somewhat unexpected, as Tuell & Isaacs (2010) suggested that social bees
might be more affected by typical blueberry pest management programs because colonies remain
active during the periods of more intensive insecticide use for pest control following crop bloom.
However, only one of three studied social species in Tuell & Isaacs (2010) supported this
expectation. Two previous meta-analyses of wild bee species’ responses to environmental
disturbance (Winfree et al., 2009; Williams et al., 2010) found that social species exhibited
significantly more negative responses to pesticide use than solitary species. However, none of
the agricultural crops included in the Williams et al. (2010) meta-analysis (pumpkin, squash,
watermelon, and olives) are comparable to highbush blueberry in terms of growth habits,
seasonality, or pest management needs. Additionally, the Winfree et al. (2009) meta-analysis
includes the same pumpkin and squash paper as Williams et al. (2010) (Shuler et al., 2005), as
well as a study of the effects of aerial applications of an organophosphate on exposure-caged
bees (Plowright & Rodd, 1980). The types of chemicals, seasonality, and routes of exposure to

67

those chemicals may differ for wild bees in the highbush blueberry system, leading to differences
in species’ responses.
Solitary species may be more likely to be negatively affected than social species when
exposed to insecticides because a solitary female bee is the sole provider for her nest and all of
her reproductive potential is therefore lost if she is killed or otherwise impaired prior to mating
or laying eggs (Brittain & Potts, 2011). In contrast, social species can rely on multiple foragers to
provide resources; if one nestmate is killed while foraging in a treated field, there may be others
foraging elsewhere in the landscape that are not affected and can continue to provide pollen and
nectar resources to the colony (Brittain & Potts, 2011). Additionally, in some socially
polymorphic species, such as Halictus ligatus Say, female workers can mate and lay fertilized
eggs, even in queenright nests; if the nest-founding queen is killed, she can be replaced by a
recently eclosed worker (Richards et al., 1995; Rehan et al., 2013). This flexibility in foraging
and egg-laying may afford social species an advantage in intensively managed agricultural areas.
In a study of the effects of different agricultural practices and landscape structure on bee
communities, Le Féon et al. (2010) found that Bombus spp. were relatively more abundant at
sites with high inputs of insecticides and nitrogen fertilizers, while solitary species were less
abundant at those sites. More work is needed to characterize the timing and identity of
insecticide applications that may be particularly deleterious to solitary or social bees in this and
other crop systems.
The observed difference between social and solitary bee species responses to IPR might
also be related to the scale at which those species respond to the landscape; social species may be
responding to insecticide risk at a larger scale than the immediate field scale. Both abundance
and richness of social species were negatively associated with the first principal component for

68

landscape, which represented increasing area of blueberry in the surrounding landscape. As the
area of managed blueberry increases in the landscape, the potential for exposure to insecticides
for bees foraging in that landscape also increases. Because social species, on average, travel
longer distances than solitary species to forage for nectar and pollen resources (Zurbuchen et al.,
2010), they may respond more strongly to exposure risk at the landscape rather than the field
scale.
These results are suggestive of wild bee declines related to intensive insecticide use in
this cropping system. However, wild bee communities, including those studied here, are
characterized by large annual variation in bee abundance, richness, and composition (Cane &
Tepedino, 2001; Roubik, 2001; Williams et al., 2001), and thorough collections at multiple time
points and multiple sites within each year are needed over a long period before declines (if
present) could be detected (Roubik, 2001; Williams et al., 2001; Lebuhn et al., 2013). In 2013,
only 421 wild bees were collected over 3 days of pan trapping at the 15 sites, fewer than half the
average number of wild bees captured each year during bloom in 2004-2006 by Tuell et al.
(2009). Had similar numbers of bees been collected in 2014, declines might have been suspected.
However, nearly double the 2013 captures were collected in 2014, indicating that low captures in
2013 may have been a single-year fluctuation in bee abundance in response to an unusually
warm spring in 2012, which was followed by frost damage to spring-blooming plants and a
severe midsummer drought.
The record-breaking heat and drought of 2011 and 2012, including July 2012, the hottest
month in the instrumental record (Karl et al., 2012), may have led to a short-term reduction in
wild bee abundance by reducing the quantity and quality of nectar and pollen resources available
in the landscape (Alarcón et al., 2008; Minckley et al., 2013), by disrupting physiological

69

development or reproductive processes (Bale & Hayward, 2010; Sgolastra et al., 2011), and/or
by desiccating developing larvae and prepupae in uninsulated environments, such as nests
located aboveground or belowground nests close to the soil surface (Cane & Neff, 2011). Adult
females foraging for pollen and nectar in summer 2012 may have had to visit more flowers and
travel longer distances to collect sufficient pollen and nectar for their developing brood; this
increase in energy expenditure during foraging is likely to reduce the number of offspring
produced by individual females (Williams & Kremen, 2007).
Fluctuations in bee populations due to these types of climatic perturbations and other
variables can make detection of overall declines challenging. Roubik (2001) reported annual
variability of bee abundance in a tropical bee community of up to 300%, or fourfold changes in
bee abundance from year to year, and annual changes in bee richness of up to 14-fold. Long-term
collections by a single researcher in the E.S. George Reserve in Michigan over several decades
corrected for differences in sampling intensity also showed wide variation in bee abundance
among years, with five times as many bees collected in each of two years in the 1970s than in
any other year (Williams et al., 2001). In addition, the sites in my study included only one
unmanaged site. To separate natural variation from an environmental impact of interest – for
example, intensive insecticide use in this system – any future sampling should include additional
unmanaged sites to account for natural community variability removed from specific disturbance
factors (Williams et al., 2001).
The diversity of individual species’ responses to environmental stressors may be an
important stabilizing mechanism for species-rich communities and, by extension, the ecosystem
services that these communities provide (Elmqvist et al., 2003; Hooper et al., 2005; Winfree &
Kremen, 2009; Winfree, 2013). Species-specific responses to insecticide risk and landscape

70

composition in this study differed widely, with some species, such as the Vaccinium oligolege
Andrena carolina, exhibiting strong negative responses to insecticide risk, while others, such as
the social halictid Lasioglossum pilosum, being positively correlated with insecticide risk. These
responses may be mediated by life-history traits, such as sociality. Maintaining a broad diversity
of species and associated functional traits may help to buffer these communities and the plants
they pollinate against environmental stressors, including intensive agricultural practices,
conversion of land for development, and changing plant phenologies due to climate change
(Winfree et al., 2007; Winfree & Kremen, 2009; Bartomeus et al., 2011; Blüthgen & Klein,
2011; Albrecht et al., 2012). Management strategies for the enhancement of pollination service
delivery in highbush blueberry should therefore place a priority on conserving functional
diversity as well as species diversity in order to improve the annual stability of these vernal
pollinator communities, as well as their resilience to current and future environmental
disturbances.
Not all of the bees captured in pan traps and nets provide pollination services to
blueberry, and their responses to farm management practices and landscape may be of less
interest to growers intending to conserve bees solely for their value as pollinators of blueberry.
However, the subset of bees providing pollination services in blueberry responded no differently
to pest management than the community as a whole. My analysis did not weight individual
species by their pollination efficiency on blueberry (Javorek et al., 2002), which might provide a
more accurate picture of how pollination services may be affected by species-specific responses
to environmental stress. However, continued intensive pest management is likely to have longterm deleterious effects on the wild bee communities living in and around blueberry fields, as
well as the pollination they provide. If growers want to increase pollination service delivery by

71

wild bee species, they will likely need to adopt approaches to pest management that minimize
risk to bees, such as spraying at night, minimizing spray drift, and selecting the least toxic
insecticide options to control target pests.

72

CHAPTER 3:

PERENNIAL WILDFLOWER PLANTINGS
SUPPORT INCREASED NESTING BY SOIL-NESTING BEES

73

Introduction
Bees (Hymenoptera: Apoidea) provide essential pollination services to wild angiosperms
and many cultivated agricultural crops (Klein et al., 2007; Winfree et al., 2007; Ollerton et al.,
2011; Garibaldi et al., 2013). However, they face complex challenges in agricultural
environments, including loss, fragmentation, and degradation of suitable habitats; exposure to
agricultural chemicals; and infection with pathogens and diseases (Kremen et al., 2002; Potts et
al., 2010; Kennedy et al., 2013; Goulson et al., 2015). Of these, the decline in foraging and
nesting resources associated with agricultural intensification may be the most fundamental cause
of observed declines in wild bee populations in managed ecosystems (Kremen et al., 2002;
Tscharntke et al., 2005; Biesmeijer et al., 2006; Goulson et al., 2008, 2015; Kleijn & Raemakers,
2008; Grixti et al., 2009; Naug, 2009; Potts et al., 2010; Winfree, 2010). There is a critical need
to conserve and enhance wild bee populations, particularly in agricultural landscapes with
pollination service demands (Allen-Wardell et al., 1998; Klein et al., 2007; Winfree, 2010), but
this requires knowledge of to what extent potential conservation strategies support bee
populations in these settings.
Strategies for conserving wild bees in agricultural landscapes have focused on mitigating
habitat loss, primarily through floral restorations such as hedgerows, wildflower strips, and
meadows (Winfree, 2010). Government subsidy programs that were developed to remove arable
land from production, such as the set aside programs under the Common Agricultural Policy of
the European Union and the USDA Conservation Reserve Program in the United States, now
include incentives for restoration projects aimed at conserving farmland biodiversity (Goulson,
2003b; Winfree, 2010). The response of bee populations to these biodiversity restorations can be
mixed (Kleijn et al., 2006; Carvell et al., 2007); however, more targeted restorations that include

74

the establishment of bee-attractive flowering plants to provide pollen and nectar resources
throughout the growing season have consistently been found to increase wild bee abundance and
richness over time (Marshall et al., 2006; Carvell et al., 2007; Pywell et al., 2011; Carvalheiro et
al., 2012; Morandin & Kremen, 2013; Blaauw & Isaacs, 2014b; Jönsson et al., 2015).
Management strategies to support bees are only as good as our knowledge of bees’
resource requirements and our ability to match those requirements with the components of the
conservation strategy. Due to the nutritional needs of both adult and immature bees, floral
resources are an important part of these requirements, and wildflower mixes are designed with
increasingly detailed consideration of bee nutritional requirements and foraging preferences
(Pywell et al., 2005; Smith et al., 2008; Tuell et al., 2008; Williams et al., 2015). However,
nesting resources may also be a limiting factor in disturbed landscapes. While the nesting
requirements and preferences of aboveground nesting bees, such as mason and leafcutter bees,
are relatively well studied (Medler, 1966; Krombein, 1967; Hubbell & Johnson, 1977; Frankie et
al., 1993; Vandenberg, 1995; Cane et al., 2007; Vickruck & Richards, 2012), little is known
about the nesting requirements and substrate preferences of most bees that nest in the soil
(Murray et al., 2009b; Winfree, 2010; but see Michener et al., 1958; Cane, 1991; Potts et al.,
2005; Cane & Neff, 2011; Sardiñas & Kremen, 2014). This paucity of nesting information
persists despite the fact that the soil-nesting guild of bees comprises around two-thirds of all
known bee species (Michener, 2000), perhaps because of the challenges in detecting the
inconspicuous nest entrances and excavating subterranean nests (Cane, 1997; Waters et al.,
2011; O’Connor et al., 2012).
Bees have evolved diverse adaptations for nesting in soil and other substrates, and
different species are likely to exhibit different ranges of tolerance for edaphic and microclimate

75

conditions (Eickwort et al., 1981; Cane, 1991; Potts & Willmer, 1997; Cane & Neff, 2011). Soildwelling sphecoid wasps, the ancestors of bees, generally do not line their nests with glandular
secretions (Michener, 1964). However, many ground-nesting bees have adaptations of this
nature, such as female cellophane bees (Hymenoptera: Colletidae), which line their nests with a
waterproof secretion from the Dufour’s gland, allowing them to live in a wide range of moisture
conditions (Michener, 1964; Hefetz, 1987; Cane, 1991). Many bees in the Halictidae and
Andrenidae families also line their nests with different combinations of water-repellent lipids
from the Dufour’s gland (Hefetz, 1987; Cane, 1991). Bees in the genus Megachile
(Hymenoptera: Megachilidae) have diverse nesting strategies, with some members of the genus
nesting in pre-existing cavities aboveground and others excavating shallow soil nests (Eickwort
et al., 1981). All megachilids, however, share the behavioral adaptation of using weatherresistant foreign materials, such as leaves, resin, or mud, to partition and line the cells of their
nests as an alternative to the glandular secretions used by other guilds (Eickwort et al., 1981).
Different tolerances or preferences for edaphic and climatic conditions mean that nestfounding females may search for specific microhabitat characteristics in choosing a nest site
(Potts & Willmer, 1997). Bumble bees (Hymenoptera: Apidae; Bombus spp.) search for large
pre-existing cavities, such as abandoned rodent burrows, to found their nests. Many other bees
also tend to reuse pre-existing cavities rather than excavate new nests (Cane, 1991). In an
observational study of roadside banks, Michener et al. (1958) found that the distribution of bee
nests in the banks was influenced by large-scale microhabitat characteristics, such as slope,
aspect, and soil type, but not by finer-scale characteristics such as the particle size, pH, or soil
color within a soil type. Several studies have also suggested that bees prefer to nest in areas with
exposed bare ground (Potts et al., 2005; Hopwood, 2008; Sardiñas & Kremen, 2014). It is

76

unclear, however, how limiting the distribution of preferred microhabitat characteristics in
human-altered landscapes may be for bee species with different ranges of edaphic and climatic
tolerances, and whether these characteristics can be used to accurately predict bee nesting in
different habitat types.
Because bees are central-place foragers that return to their nest after each foraging bout,
nest location is an essential piece of information needed to predict the distribution of bees and
their associated pollination services across the landscape (Lonsdorf et al., 2009; Sardiñas &
Kremen, 2014). Lonsdorf et al. (2009) developed a spatial model of native bee abundance in
agricultural landscapes based on nesting resource availability and the forage resources within
their flight range, which can be used for predicting wild bee contributions to crop pollination
across the landscape and developing landscape-scale conservation strategies (Lonsdorf et al.,
2009, 2011; Kennedy et al., 2013). However, the dearth of empirical information on bee nesting
preferences has necessitated the use of expert opinion surveys to determine the nesting suitability
of different land use/land cover classes, a key model parameter (Lonsdorf et al., 2009; Chapin,
2014). These expert opinions can differ widely among people, leading to high uncertainty for the
nesting parameter within the models (Murray et al., 2009a; Chapin, 2014). This uncertainty
increases the model prediction error and decreases the correlation with observed patterns of bee
abundance, particularly as the Lonsdorf et al. (2009) model has been found to be sensitive to
altering the parameters for soil-nesting bees (Chapin, 2014). Better information about the nesting
suitability of different habitats and the mechanisms underlying bee nesting preferences is needed
to improve the ability to use these models to inform bee conservation and management decisions.
For this study, I chose to focus on the pollination-dependent blueberry agroecosystem of
southwest Michigan. Highbush blueberry (Vaccinium corymbosum L.), the main blueberry

77

species grown in the state, is highly dependent on insect-mediated pollination for economically
viable yields (Merrill, 1936; Schaub & Baver, 1942; Meader & Darrow, 1947; Free, 1993), and
at least 80% of flowers must set fruit to achieve a commercial-scale blueberry crop (Merrill,
1936). Wild bees are important contributors to pollination of this crop, supplying approximately
12% of the yield increase due to pollination in large commercial fields and nearly all of the
pollination (88%) in smaller, more isolated fields that are not stocked with managed honey bee
(Apis mellifera L.) hives for pollination (Isaacs & Kirk, 2010). The cost associated with renting
hives, and the concerns about future health and availability of this managed species, has led to
considerable interest in on-farm strategies to increase wild bee contributions to pollination.
Current approaches include planting wildflowers in fallow areas or field margins and providing
nesting materials such as wood blocks or cardboard tubes for mason bees; however, little is
known about the preferred nesting resources of most soil-nesting bee species, which comprise
most of the wild species that supply pollination in this region (Tuell et al., 2009). Understanding
where bees choose to nest in this farm landscape could help to determine which areas are most
suitable for soil-nesting bees, and how to best mitigate risk to nesting bees from management
practices such as tillage or pesticide use.
The main objectives of this study were 1) to evaluate the nesting suitability of different
land use types and associated microhabitat characteristics for soil-nesting bees on blueberry
farms, and 2) to determine whether perennial wildflower restorations enhance nesting by soilnesting bees, and 3) to determine whether restoration age affects the abundance of nesting bees.
For the first objective, I sampled soil-nesting bees in four different land use types on blueberry
farms that included established perennial wildflower plantings next to managed crop fields. For
the second and third objectives, I compared the abundance and diversity of soil-nesting bees

78

nesting in three newly-established wildflower plantings, three mature plantings, and unrestored
areas in the same landscapes.

Materials and Methods
Nesting suitability of blueberry farm habitats
Study sites
This study was conducted at four blueberry farms located in Allegan, Berrien, and Van
Buren counties in southwest Michigan with mature wildflower plantings sown with a diverse
mix of native forbs and grasses in 2008 or 2009 (Figure 3.1). These wildflower plantings ranged
in size from 0.3 to 3 acres. At each farm, mesh emergence traps (60cm2; MegaView Science,
Taiwan) were used to sample soil-nesting bees in four different habitat types: the blueberry crop
field, an adjacent grassy field margin, a wooded area, and the wildflower planting. The crop
field, field margin, and wooded areas selected for trapping were located at least 100m from the
wildflower planting at each site to reduce potential spillover from the plantings (Figure 3.2).
These habitats represent the dominant land use types in the Michigan highbush blueberry
agroecosystem, and were present at all sites.

Emergence trapping
To sample soil-nesting bees, ten traps were placed at 5m intervals along a haphazardlyplaced 50m transect in each of the four habitats - the crop field, field margin, wooded area, and
the wildflower planting - for a total of 40 traps per site per sample round. Traps were placed at
midday (between 0900 and 1500 hours) and left for 2-3 days prior to collection. Bees emerging
from under a trap were captured in a jar containing 2% soap solution at the apex of the trap

79

(Dawn Ultra Original Scent, Procter & Gamble Co., Cincinnati, OH). Specimens were strained
into ziplock bags and frozen before being washed, dried, and identified. This process was
repeated three times per farm from late June to early September in 2013 and 2014, for a total of
six sampling events per farm. Pesticide applications in the crop field interfered with trap
placement in several sample rounds, resulting in uneven sampling in the crop field relative to the
three other habitats, with an average of 3.2 sample rounds per site in the crop field over the two
years of sampling.

Figure 3.1. Map of study sites sampled for soil nesting bees. Dark circles represent the
approximate locations of four blueberry farms sampled using emergence traps in four different
habitat types per farm in 2013 and 2014. Light rectangles represent newly-established wildflower
restorations sampled in 2014. Light triangles represent mature wildflower plantings sampled in
2014. Each restoration site had a paired unrestored control site within 400 m (not marked).

80

Woods!
Wildflower!
planting!
Blueberry!
crop!field!

Field!
margin!

Figure 3.2. Representative distribution of habitat types and sampling locations at one of four
blueberry sites sampled using emergence traps and netting in 2013 and 2014. Imagery © 2015
GoogleTM.

!

Bee species richness was characterized for emergence trapped bees pooled by habitat by
farm using the Chao1 abundance-based richness estimator in the estimateR function of R
package vegan, which corrects for undetected species (R version 3.0.2, R Core Team, 2013;
Chao, 1987; Oksanen et al., 2013). Non-parametric Kruskal-Wallis tests were used to determine
whether bee abundance, richness, or Simpson’s (1/D) diversity in the emergence traps differed
among the four habitats, as these data violated the assumption of normality for analysis of
variance. The abundance variable included all trapping events (n=838), while Chao1 estimated
richness and Simpson’s (1/D) diversity were calculated by habitat by farm (n=16). Means
separation for abundance of bees in different habitats was carried out using a post-hoc Dunn’s
test with Bonferroni correction for p-values for the multiple comparisons among habitat types,

81

which as a test for median difference is an appropriate multiple comparisons method for
unbalanced designs (Dunn, 1961, 1964).
Microhabitat characteristics
Prior to placing a trap, I recorded microhabitat characteristics in the 60cm2 sampling
quadrat according to the methods of Potts et al. (2005) and Sardiñas and Kremen (2014) (Table
3.1). Percent cover attributes did not always sum to 100%, as all surface-level categories (bare
ground, leaf litter/thatch, rocks, and wood) were counted as a percentage of the quadrat area
regardless of tall vegetation. Surface cracks were measured as a percentage of the quadrat area,
but were nested within bare ground. Surface soil strength (kg/cm2) was measured in three
locations in each quadrat using a pocket soil penetrometer (Forestry Supplies, Inc., Jackson,
MS). Following measurement of microhabitat characteristics but prior to trap placement, surface
vegetation was cut back to a height of 4-6 inches using manual hedge shears and the area was
examined thoroughly to ensure no bees were visible above the surface.

Table 3.1. Microhabitat characteristics measured in each emergence trap quadrat in four habitat
types on blueberry farms in 2013 and 2014.
Microhabitat characteristic
Unit of measurement
Vegetation
% cover (to nearest 1%)
Bare ground
% quadrat area (to nearest 1%)
Leaf litter/thatch
% quadrat area (to nearest 1%)
Rocks
% quadrat area (to nearest 1%)
Wood (including sticks)
% quadrat area (to nearest 1%)
Surface cracks
% quadrat area (to nearest 1%)
Cavities
# of cavities of <2 cm or >2 cm diameter
Ant nests
# of ant nests
Shade
Full, partial, or none
Slope
To nearest 1%
Aspect
N, NW, W, SW, S, SE, E, NE
Soil surface compaction
Three penetrometer readings

82

Microhabitat measurements were assessed for collinearity using a Spearman’s rank
correlation matrix (Table 3.4). To address the issue of multicollinearity in these data, I used
principal components analysis to reduce the dimensionality of the microhabitat variables
(function prcomp in R package stats; R Core Team, 2013). Visual examination of the scree plot
supported the extraction of the first two principal components for use as predictor variables in a
model with bee abundance in emergence traps as the response variable (Figure 3.3; Table 3.5).

Figure 3.3. Scree plot of the percentage of variation explained by the principal components
generated from 11 microhabitat characteristics measured in emergence traps in four different
habitat types on blueberry farms in 2013 and 2014 (n=645 traps).

83

Because the bee capture data included many zeros, the data were modeled using a zeroinflated regression model with Poisson distribution (function zeroinfl in R package pscl; Zeileis
et al., 2008), which includes a point mass at zero and is typically better suited for modeling overdispersed count data than a generalized linear model (GLM) with Poisson distribution. Model
selection for the two principal components was conducted using backward stepwise elimination
based on Akaike’s information criterion (AIC) values (function step in R package stats; R Core
Team, 2013). To ensure optimal model fit, I compared the zero-inflated Poisson model with a
zero-inflated negative binomial regression model using AIC values, and with a GLM with
Poisson distribution using Vuong’s likelihood ratio test (function vuong in R package pscl;
Vuong, 1989). The zero-inflated regression model with Poisson distribution exhibited the best fit
of the competing models.
Visual examination of the relationship between the first two principal components and
habitat type suggested that the microhabitat characteristics vary non-independently from habitat
type (Figure 3.6). For example, on average, crop fields had more bare ground than other habitats,
which is expected given that most fields were treated with herbicides for weed control under the
blueberry bushes. Analysis of variance (ANOVA) was used to determine whether the principal
components for microhabitat differed significantly by habitat type. The second principal
component was log(x+5) transformed prior to analysis to improve normality. Means separation
by habitat was performed using Tukey’s HSD (α = 0.05).
Following the analysis of the first two principal components, I used logistic regression to
explicitly examine the relationship between presence and absence of bees in traps and the
microhabitat variables with factor loadings greater than 0.4 (Table 3.5). Logistic regression
allows for the direct derivation of the change in probability of encountering a bee in a trap given

84

a unit increase in a predictor variable. Because percent bare ground and number of small cavities
were positively correlated (Table 3.4), each variable was first analyzed in separate logit models
of bee presence in traps. These individual models were then compared with the combined model
using AIC values (Table 3.5). The model coefficients (β) remained stable in the combined
model, indicating that the positive correlation between the two predictor variables did not lead to
problems with collinearity in the combined model. Wald’s χ2 tests were used to determine the
significance of the individual model coefficients (Table 3.6).

Soil characteristics
To measure the compaction and texture of the topsoil, I took 10 replicate soil samples per
habitat per farm in July 2013 using a manual drop hammer soil core sampler to a depth of 7.5 cm
(331.3 cm3 core volume). In order to calculate soil moisture and bulk density, fresh soil samples
were placed in #5 paper bags and weighed, then oven dried at 105°C for 24 hours. Dried samples
were set aside to cool for 15 minutes prior to measuring dry weights. Soil mass was determined
by subtracting the mean weight of ten empty oven-dried #5 bags from the dried sample weights.
Samples from each habitat were then processed through a 2mm mesh sieve before being
aggregated and sub-sampled for texture. Bulk density was calculated as follows (Robertson et
al., 1999):
Bulk density (g/cm3) = Dry soil weight (g) / Soil volume (cm3)
Particle size analysis for soil texture determination was conducted using a hydrometer
method adapted from Day (1965). From each habitat, 40 g +/- 0.05 g of sieved soil was weighed
into a 150mL shaker bottle, to which 100mL of a 5% sodium hexametaphosphate dispersing
solution was added. Bottles were capped and placed on a reciprocating horizontal shaker at 150

85

rpm for two hours. After two hours, the suspension was transferred to a 1.0 L sedimentation
cylinder, and deionized water was added to bring a 1.0 L final volume. The suspension was
allowed to equilibrate to room temperature for two hours before being mixed thoroughly with a
plunger. After 30 seconds, a hydrometer (ASTM No. 1. 152H-Type with Bouyoucos scale in g L1

) was lowered into the suspension, and at 40 seconds a reading was taken to the nearest 0.5 g L-1

(reading “Rsand”). This process was repeated with a blank solution (reading “RC1”). The soil
suspensions were left for six hours, after which the temperature of the suspension was recorded.
A second hydrometer reading was taken at the settling time for clay, as determined by the
temperature of the suspension (reading “RClay” or “RC2” for the blank solution). Oven dry soil
moisture was determined on a separate 40 g sample of soil from each habitat. Soil texture was
calculated as follows:
Sand % = ((oven dry soil mass – (Rsand – RC1))/(oven dry soil mass) x 100
Clay % = (Rclay – RC2)/(oven dry soil mass) x 100
Silt % = 100 – (Sand % + Clay %)
A non-parametric Kruskal-Wallis test with posthoc Dunn’s test for pairwise multiple
comparisons was used to determine whether bulk density (g/cm3) differed among habitats.
Kruskal-Wallis tests were used to determine whether average bulk density by habitat was
correlated with bee abundance and richness in emergence traps.
Analyses of variance (ANOVAs) with post-hoc Tukey’s HSD for multiple comparisons
were used to determine whether soil texture characteristics differed by site or habitat. Percent silt
and percent clay data were square root transformed prior to analysis to improve normality.
Significance levels for these tests were adjusted with the Bonferroni correction.

86

The three soil texture characteristics were strongly intercorrelated (Table 3.11), so
principal components analysis was used to reduce the dimensionality of the texture data while
preventing issues with multicollinearity (function prcomp in R package stats; R Core Team,
2013). The first principal component explained over 95% of the variance in the soil texture data
and was retained for use in a simple linear regression with bee abundance in traps (Table 3.12).
Average bee abundance in each habitat, the dependent variable in the regression, was log(x + 1)
transformed prior to analysis to improve normality.

Bee communities collected in nets vs. emergence traps
To determine how the community of soil nesting bees compared with the flower-visiting
bee community, I net collected bees visiting open flowers for 10 minutes in 2013 and 20 minutes
in 2014 along the 50m transect in each habitat type during suitable weather conditions
(temperature >65oF, wind speed <3.5m/s) for each soil emergence trapping session. Because
emergence traps were placed regardless of weather conditions, net collections were sometimes
conducted 1-3 days after trap placement.
To visualize differences in the bee communities captured in emergence traps and netcollected from open flowers, I used non-metric multidimensional scaling (NMDS) ordination
using Bray-Curtis dissimilarities with Wisconsin double standardization (functions wisconsin
and metaMDS, R package vegan). Bee communities were pooled across the two years of
sampling for each site. Because of the low abundances of bees in emergence trap samples by site,
Bray-Curtis dissimilarities were zero-adjusted using a dummy species with a value of 1 for all
sites in the original abundance matrix (Clarke et al., 2006). Permutational multivariate analysis
of variance (perMANOVA) based on Bray-Curtis dissimilarities with 1000 permutations was

87

used to compare netted and trapped bee communities (function adonis, R package vegan;
Anderson, 2001; McArdle & Anderson, 2001; Oksanen, 2006). Multivariate homogeneity of
group dispersions (variances) was confirmed using function betadisper in R package vegan, a
multivariate analogue of Levene’s test (Anderson, 2006).

Bee identification
Bee specimens (Hymenoptera: Apoidea) were identified to genus and species using published
dichotomous keys (Mitchell, 1960, 1962; LaBerge, 1980; McGinley, 1986; Michener et al.,
1994; Gibbs, 2011; Rehan & Sheffield, 2011) and the online keys available through
www.discoverlife.org. Emergence trap specimens and all Lasioglossum (Dialictus) and Andrena
specimens were identified by Jason Gibbs (Department of Entomology, Michigan State
University). Voucher specimens from this study were deposited in the Albert J. Cook Arthropod
Research Collection at Michigan State University.

Suitability of new and mature wildflower restorations as nesting sites for bees
Study sites
This study was conducted in three newly-sown and three mature wildflower restorations in
southwest Michigan. Each restoration was paired with an unrestored control habitat within 400 m
in order to compare nesting between sites with similar landscape composition and environmental
conditions with and without the addition of wildflowers. New restorations were sown into
herbicide-treated bare ground adjacent to blueberry fields in fall 2013 or spring 2014 with a
mixture of 26 forb species and four grasses. These ranged in size from approximately 0.3 to 0.8
acres. Control habitats for new restorations were grass-dominated blueberry field margins, two of

88

which received regular mowing in summer 2014. These were located between 135 and 350
meters away from the restoration. Mature restorations ranged in age from five to ten years, in
size from 4 to 144 acres, and in sown floristic diversity from 22 to 48 forb species. Unrestored
control sites for mature restorations were unmowed old field habitats with varying levels of floral
diversity, which ranged in distance from approximately 30 to 185 meters from the edge of the
wildflower restorations. New and mature restorations were spaced at least 25 km apart (Figure
3.1).

Emergence trapping
In each sample round, ten emergence traps as described above were placed at 5m intervals
along each of two haphazardly located 50m transects in the unrestored control and the
wildflower planting, for a total of 20 traps per habitat type. Traps were placed after dusk
(between 20:00 and 22:00 hours) to ensure that bees were resident in their soil nest during trap
placement, and were left for one week prior to collection. Sample rounds were conducted in the
same week for one new and one mature restoration and their respective paired unrestored
controls, with three sample rounds between June-August 2014.
Bee species richness was characterized for trapped bees using the Chao1 abundance-based
richness estimator in the estimateR function of R package vegan, which corrects for undetected
species (Chao, 1987; Oksanen et al., 2013). The abundance, richness, and Simpson’s (1/D)
diversity of bees captured in emergence traps were compared between habitats using
nonparametric Kruskal-Wallis tests, as these response variables did not meet the assumption of
normality. The abundance data used in this test were bee counts from all emergence traps
(n=720), while Chao1 estimated richness and Simpson’s diversity were calculated by habitat by

89

site (n=12). Means separation was performed using post-hoc Dunn’s tests with Bonferroni
correction.

Microhabitat characteristics
Prior to placing each trap, I recorded basic surface characteristics in the trap quadrat (Table
3.2) and cut vegetation to a height of 4-6 inches using manual hedge shears, as described above.
Collection jars were half-filled with a 2% soap solution as described above. Collected specimens
were strained, then funneled into a 50ml polypropylene vial (Corning Inc., Corning, NY). Vials
were filled with 70% ethanol for storage. In the laboratory, bee specimens were sorted out of the
vials before being washed, dried, and identified to species. All specimens for this study were
identified by Jason Gibbs (Department of Entomology, Michigan State University) using the
keys noted above.
Microhabitat characteristics were examined for multicollinearity using a Spearman’s rank
correlation matrix (Table 3.15). The variables were highly intercorrelated, so I used principal
components analysis to reduce the dimensionality of the variables without creating problems
with multicollinearity (function prcomp in R package stats; R Core Team, 2013). Visual
examination of the scree plot supported the extraction of the first four principal components for
use as predictor variables in a model of bee abundance in traps (Figure 3.4; Table 3.16).
Together these four components accounted for 96.6% of the variation in the microhabitat
variables.

90

Table 3.2. Microhabitat characteristics measured in each emergence trap quadrat in three newlyestablished wildflower plantings, three mature wildflower plantings, and paired unrestored
controls in 2014.
Microhabitat characteristic
Unit of measurement
Vegetation
% cover (to nearest 1%)
Bare ground
% quadrat area (to nearest 1%)
Leaf litter/thatch
% quadrat area (to nearest 1%)
Wood/rocks
% quadrat area (to nearest 1%)
Slope
Flat, 1-5%, 5-10%, 10-25%, 25-50%, >50%
Aspect
N, NW, W, SW, S, SE, E, NE

Figure 3.4. Scree plot of the percentage of variation explained by the 5 principal components
generated from the microhabitat data from emergence trap quadrats in new wildflower
restorations, mature wildflower restorations, and their paired unrestored controls in 2014 (n=720
traps).

91

As with the previous emergence trap data, bee counts in emergence traps were highly
overdispersed. In addition, several traps captured multiple members of the same social species,
leading to outliers in the raw abundance data (Table 3.13). Thus, to analyze the effect of
microhabitat characteristics on bee nesting, I used binary presence/absence data as the response
variable in a logistic regression. Model selection with the four principal components as predictor
variables and bee presence/absence in emergence traps as the response variable was conducted
using backward stepwise elimination.

Bee community in nets vs. emergence traps
Flower-visiting bees were net collected for 50 minutes in the wildflower planting or
unrestored control within 1-2 days of trap placement (20 minutes along each of the two 50m
transects and 10 minutes of opportunistic collection). In addition, I counted the number of open
flowers of each flowering species in 20 1m2 quadrats placed at 5m intervals along each 50m
transect. Community analyses comparing netted bees with emergence trapped bees were
conducted as above. The contribution of individual bee species to the contrast between trapped
and netted bee communities was assessed using similarity percentage (SIMPER) analysis
(function simper in R package vegan; Clarke, 1993), which conducts pairwise comparisons of
groups of sampling units and finds the average contributions of each species to the overall BrayCurtis dissimilarity.

92

Results
Nesting suitability of different habitats on blueberry farms
I captured a total of 47 soil-nesting bees in 42 traps in the four blueberry farm habitats out
of a total of 838 trapping events over two years (Table 3.3). Captured bees included specimens
from four of five bee families present in Michigan, but were primarily composed of species in
the Halictidae and Apidae families (Table 3.3). Four cleptoparasite species were captured in
traps: Nomada illinoensis Robertson, Nomada vegana Cockerell, Sphecodes davisii Robertson,
and Sphecodes mandibularis Cresson. Four species of stem-nesting bees (Ceratina spp. and
Hylaeus affinis) were also captured in emergence traps.

93

Table 3.3. Bee species collected in emergence traps from different habitat types on four
blueberry farms in 2013 and 2014. No. traps represents the number of individual traps (and
therefore the minimum number of possible nests) from which these species were collected.
Location captured

Family
Species
Andrenidae
Andrena canadensis
Andrena miserabilis
Apidae
Anthophora bomboides
Bombus bimaculatus
Bombus griseocollis
Bombus impatiens
Ceratina calcarata*
Ceratina dupla*
Ceratina strenua*
Nomada illinoensis**
Nomada vegana**
Colletidae
Colletes thoracicus
Hylaeus affinis*
Halictidae
Agapostemon sericeus
Agapostemon splendens
Augochlorella aurata
Lasioglossum ellisiae
Lasioglossum fuscipenne
Lasioglossum leucocomum
Lasioglossum macoupinense
Lasioglossum pectorale
Lasioglossum pilosum
Lasioglossum versans
Lasioglossum vierecki
Sphecodes davisii**
Sphecodes mandibularis**
Total:

Woods

Total no.
captured

No.
traps

3
1

0
0

3
1

1
1

0
2
0
3
0
0
0
0
0

0
0
0
0
0
1
0
0
0

0
1
0
0
1
0
0
1
0

1
3
1
3
1
1
1
1
1

1
3
1
3
1
1
1
1
1

0
0

1
1

0
0

0
0

1
1

1
1

0
0
0
1
0
2
1
0
0
0
0
1
0

3
0
2
1
0
2
0
1
4
0
2
1
1

1
1
0
1
0
0
0
0
1
0
0
0
0

0
0
0
0
1
0
0
0
0
1
0
0
0

4
1
2
3
1
4
1
1
5
1
2
2
1

3
1
2
3
1
3
1
1
4
1
2
2
1

9

24

9

5

47

42

Crop
field

Flower
planting

Field
margin

0
0

0
0

1
0
1
0
0
0
1
0
1

* Stem-nesting bees
** Cleptoparasitic bees

94

The abundance of soil-nesting bees differed significantly among habitats (Figure 3.7; χ2 =
13.56, df = 3, p = 0.003), with the greatest captures in the wildflower plantings compared to
those in the wooded areas (posthoc Dunn’s test for multiple comparisons with Bonferroni
correction, p = 0.01) and grassy field margins (p = 0.03). Crop fields had an intermediate
abundance of soil-nesting bees, but did not have significantly fewer bees than wildflower
plantings or more than the other two habitats. Bee species richness in emergence traps did not
differ significantly among habitats (χ2 = 5.62, df = 3, p = 0.13), nor did Simpson’s (1/D)
diversity of the trapped bees (χ2 = 3.09, df = 3, p = 0.38), though the sample size for each habitat
was relatively low.

Figure 3.5. Mean number of bees captured in emergence traps in different habitat types on four
blueberry farms in 2013 and 2014. Captured bee abundance does not differ significantly among
habitats marked with the same letter (posthoc Dunn’s pairwise comparisons with Bonferroni
correction, α=0.05).
95

Microhabitat characteristics
Many of the microhabitat characteristics were strongly intercorrelated (Table 3.4),
particularly among the percent cover categories. Living vegetation, the dominant percent cover
category, was negatively correlated with most of the other cover types. Percent bare ground was
strongly negatively correlated with the percent cover of leaf litter and thatch, but positively
correlated with the number of cavities and ant nests.
Microhabitat factors loading strongly on the first principal component were leaf litter and
thatch cover, which loaded positively, and soil strength (kg/cm2) and living vegetation cover,
which loaded negatively (Table 3.5). Bare ground and the number of ant nests also had weak
negative loadings on PC1. Bare ground and the number of small cavities had strong positive
loadings on the second principal component, along with weak positive loadings from the number
of large cavities and slope and a weak negative loading from vegetation cover (Table 3.5).
The two principal components exhibited nearly equal but opposite relationships with bee
abundance in traps, with bee abundance exhibiting a negative relationship with leaf litter and
thatch cover (PC1) and a positive relationship with bare ground and the number of small cavities
(PC2) (Table 3.6). AIC values for the models containing either one of the two principal
components and the model with both components were nearly equal; all models were within 1
AIC value of one another (Table 3.6). Given that bare ground was negatively correlated with
vegetation and litter (Table 3.4), these axes had similar explanatory power for bee abundance.
Likelihood ratio tests indicated that all three models were significantly better than the null model
at α = 0.05 (Table 3.6).

96

Table 3.4. Spearman’s rank correlation matrix of 11 microhabitat characteristics measured in each 60cm2 emergence trap quadrat on
four blueberry farms in 2013 and 2014 (n=645 traps).
Microhabitat
Small
Large Ant
Soil
characteristic
Bare
Vegetation Rocks Litter Wood Cracks cavities cavities nests strength Slope
Bare ground (%)
1
Vegetation (%)
-0.17***
1
*
Rocks (%)
0.09
-0.02
1
Leaf litter/thatch (%)
-0.56***
-0.35***
-0.08*
1
***
***
Wood (%)
-0.22
-0.54
-0.01
0.38***
1
***
Cracked surface (%)
0.02
-0.13
0.03
0.09* 0.13***
1
***
*
*
No. cavities (<2mm)
0.22
-0.08
-0.02
-0.09
-0.03
0.04
1
***
***
No. cavities (>2 mm)
0.16
-0.18
-0.03
0.01
0
0.07
0.16***
1
***
***
***
No. ant nests
0.34
0.05
-0.04 -0.26
-0.25
-0.06
0.04
-0.02
1
2
***
**
***
***
***
Soil strength (kg/cm )
-0.03
0.35
0.11
-0.21
-0.16
-0.04
0.01
-0.16
0.08*
1
***
*
Slope (%)
0.05
-0.04
-0.04
0.05
0.13
0.01
0.1
0.05
-0.06
-0.15***
1
*significant at p = 0.05, **p = 0.01, ***p = 0.001!

97

Table 3.5. Factor loadings for the first two principal components from the microhabitat
characteristics analysis. Variables with factor loadings greater than 0.4 are in bold.
Microhabitat
PC1
PC2
characteristic
Bare ground (%)
-0.25
0.53
Vegetation (%)
-0.44
-0.39
Leaf litter/thatch (%)
0.54
-0.18
Rocks (%)
-0.11
0.06
Wood (%)
0.32
0.05
Surface cracks (%)
0.11
0.19
No. small cavities
-0.12
0.41
No. large cavities
0.12
0.34
No. ant nests
-0.36
0.22
2
Soil strength (kg/cm )
-0.40
-0.20
Slope (%)
0.09
0.35

Table 3.6. Model coefficients (β), z, p, and Akaike’s information criterion (AIC) values
for the competing zero-inflated models of bee abundance in traps, and the p-values from
likelihood ratio tests compared with a null model.
p-value from
z
p
AIC
likelihood
β ± SE
Model
ratio test
Combined model (PC1+ PC2)
0.049*
PC1
-0.18 ± 0.13 -1.33 0.183 237.63
0.181
PC2
0.14 ± 0.10 1.34
Individual models
0.035*
PC1
-0.26 ± 0.11 -2.31 0.021* 237.35
0.039*
PC2
0.19 ± 0.09 2.23 0.026* 237.56
*significant at α = 0.05!
Table 3.7. Mean values by habitat for the microhabitat characteristics with large factor
loadings on PC1 and PC2. Values within a column followed by the same letter are not
significantly different (posthoc Dunn’s pairwise comparisons with Bonferroni correction,
α=0.05).
Habitat
Crop
Flower
Grass
Woods
"

No.
traps
128
240
238
230

Vegetation
(%)

Leaf litter or
thatch (%)

Bare
ground (%)

No. small
cavities

Soil strength
(kg/cm2)

32.3 ± 2.9 b
83.2 ± 1.3 a
80.8 ± 1.2 a
37.6 ± 1.9 b

35.4 ± 2.9 bc
37.3 ± 2.0 b
27.5 ± 1.6 c
84.7 ± 1.6 a

31.5 ± 2.9 a
9.0 ± 1.0 c
12.4 ± 1.1 b
2.8 ± 0.7 d

0.32 ± 0.07 a
0.13 ± 0.04 b
0.05 ± 0.02 bc
0.02 ± 0.01 c

0.46 ± 0.04 c
0.75 ± 0.03 b
1.27 ± 0.06 a
0.59 ± 0.02 c

98

Microhabitat characteristics loading on the first two principal components were not
independent of habitat (PC1: F = 229.6, df = 3, 641, p < 0.001. PC2: F = 55.5, df = 3,
641, p < 0.001; Table 3.7; Figure 3.6). For the first principal component, which was
positively associated with leaf litter and thatch cover and negatively associated with
living vegetation cover, all habitat types were significantly different, with woods and
crop fields on the positive side of the axis and wildflower plantings and grassy field
margins on the negative side of the axis (all p < 0.001; Figure 3.6). Crop fields fell
significantly higher on PC2, which is positively associated with bare ground and the
number of small cavities, than the other three habitats (all p < 0.001), which did not
differ from each other.
All of the individual microhabitat characteristics that loaded heavily (factor loading
> 0.4) on the first two principal components varied significantly by habitat type (Table
3.7). On the first principal component, percent cover of living vegetation was
significantly different by habitat (χ2 = 371.0, df = 3, p < 0.001), with significantly more
vegetation cover in wildflower plantings and grassy field margins than in crop fields or
woods. Leaf litter/thatch cover varied significantly among habitats (χ2 = 331.1, df = 3, p <
0.001), with more litter in woods than other habitats. Soil surface strength (kg/cm2) also
differed significantly among habitats (F = 62.19, df = 3, 382, p < 0.001). Grassy field
margins had the highest soil surface compaction, followed by wildflower plantings,
wooded areas, and crop fields, in that order (Table 3.7). On the second principal
component, the percent bare ground in emergence trap quadrats differed significantly by
habitat (χ2 = 207.0, df = 3, p < 0.001). Crop fields had the most bare ground per quadrat,
followed by field margins, wildflower plantings, and woods (Table 3.7). Finally, the

99

habitat types also differed in the number of small cavities measured per quadrat (χ2 =
32.2, df = 3, p < 0.001); on average, more cavities were counted in crop fields than in
other habitats. Wildflower plantings also had significantly more small cavities than
woods (p = 0.02), but not grassy field margins.
None of the variables with large factor loadings on the first principal component
were significant predictors of nesting bee presence in logistic regression models (Table
3.8). However, both variables loading heavily on the second principal component, the
percent bare ground and the number of small cavities, were significant predictors of
nesting bee presence (Tables 3.8 and 3.9). The logit model containing both variables
(percent bare ground + number of small cavities) exhibited the best fit of the tested
models (Table 3.8). The combined model indicated that the number of small cavities in a
trap quadrat was a better predictor of trap capture than the percent bare ground; with
every additional small cavity counted in a quadrat, the odds of capturing a bee increased
by a factor of 1.96 (95% confidence interval: 1.37 - 2.84; Table 3.9). Bare ground
exhibited a lower odds ratio than the number of small cavities; with every 1% increase in
bare ground in a quadrat, the odds of bee capture increased by a factor of 1.01 (95%
confidence interval: 1.00 – 1.03).

100

Figure 3.6. Biplot of first two principal components in the microhabitat PCA. Each dot
represents an individual emergence trap quadrat (n=645 traps). Ellipses represent 95%
confidence around the group mean for each habitat type.

101

Table 3.8. Model coefficients (β), z- and p-values for parameters, residual deviances,
degrees of freedom, and Akaike’s information criterion (AIC) values for the combined
and individual logistic regression models of microhabitat variables as predictors of bee
presence in emergence traps.
Model

z

β ± SE

p

Residual
deviance

df

AIC

Combined model
3.7
<0.001
small cavities
0.68 ± 0.18
295.59
833 301.59
2.21
0.03
bare ground (%)
0.01 ± 0.01
Individual models
4.28 <0.001
299.79
834 303.79
small cavities
0.76 ± 0.18
2.99
0.003
307.7
834
311.7
bare ground (%)
0.02 ± 0.01
0.17
313.32
834 317.32
leaf litter (%)
-0.01 ± 0.004 -1.36
0.53
314.84
834 318.84
vegetation (%)
-0.002 ± 0.004 -0.63
2
-0.89
0.37
204.17
643 208.17!
soil strength (kg/cm ) -0.34 ± 0.38
!
Soil strength model AIC value not directly comparable (measured on subset of quadrat
values)
"
Table 3.9. Logistic regression model assessing the number of small cavities and percent
bare ground as predictors of the presence of bees in emergence traps.
Microhabitat
characteristic
Small cavities
Bare ground (%)
"

Odds ratio
(95% confidence)
1.96 (1.37-2.84)
1.01 (1.00-1.03)

β ± SE

Wald χ 2

k

p

0.68 ± 0.18
0.01 ± 0.01

256.4
13.7

1
1

< 0.0001
0.0002

Soil characteristics
Bulk density (g/cm3) of surface soil differed significantly among habitats (F =
91.12, df = 3, 155, p < 0.001), with significantly higher average bulk density in grassy
field margins and flower plantings than in crop fields or wooded areas (Tukey’s HSD, all
p < 0.001; Table 3.10). Soil in crop fields was significantly more dense than soil in
wooded areas (p < 0.001). However, mean bulk density was not significantly correlated
with bee abundance or richness in emergence traps (Fabundance = 2.76, df = 1,14, p = 0.12;
Frichness = 1.67, df = 1,14, p = 0.22).

102

All soils were categorized as sands or loamy sands, with percent sand ranging from
80.0-95.6% across the sixteen samples (Table 3.10). Percent sand and silt did not differ
among sites or habitats (all p-values > 0.35). Percent clay differed significantly among
habitats (F = 6.34, df = 3, p = 0.008), with significantly more clay in soils sampled from
crop fields and wildflower plantings than from field margins (Table 3.10). Mean bee
abundance and Chao1 estimated richness by habitat were not related to the soil texture
principal component (Fabundance = 1.70, df = 1,14, p = 0.21; Frichness = 0.41, df = 1,14, p =
0.53; Table 3.12).
Table 3.10. Soil physical properties in the crop field, grassy field margin, wildflower
restoration, and wooded areas on four blueberry farms in SW Michigan.
Soil texture (%)
Bulk density Soil strength
Habitat
(g/cm3) ± SE (kg/cm2) ± SE
Sand ± SE
Clay ± SE
Silt ± SE
Crop field
91.2 (1.7) a
3.1 (0.4) a
5.6 (1.5) a
1.29 (0.07) b
0.31 (0.04) d
Field margin
94.4 (0.0) a
1.3 (0.0) b
4.4 (0.0) a
1.60 (0.04) a
1.54 (0.05) a
Wildflower
89.8 (3.8) a
3.1 (0.8) a
7.0 (3.1) a
1.56 (0.03) a
0.91 (0.09) b
Woods
90.0 (2.1) a
1.2 (0.1) ab 8.8 (2.1) a
0.94 (0.09) c
0.59 (0.04) c
Values are means ± 1 SE (in parentheses). Values within a column with the same letter
are not significantly different (p = 0.05).!
Table 3.11. Spearman’s rank correlation matrix of soil texture characteristics measured
on soil samples collected across different habitat types on four blueberry farms in 2013
(n=16).
Sand (%)
Silt (%)
Clay (%)
1
Sand (%)
-0.98***
1
Silt (%)
-0.54*
0.47
1
Clay (%)
*significant at p = 0.05, **p = 0.01, ***p = 0.001
Table 3.12. Factor loadings on the first principal component for soil texture.
Factor loading on PC1
Sand (%)
-0.75
Silt (%)
0.65
Clay (%)
0.10

103

Bee communities collected in nets vs. emergence traps
There was a marginally significant difference in the bee communities captured by
the two collection methods (perMANOVA; F = 1.63, df = 1, 6, p = 0.06). NMDS
ordination with 95% confidence ellipses around the group means for emergence trapped
bee communities and netted bee communities showed visual separation of the two
collection methods (Figure 3.7).

Figure 3.7. Non-metric multidimensional scaling of the community of bees captured in
emergence traps (etrap, black circles) and netted on open flowers (net, gray triangles) at
four blueberry farms in SW Michigan. Ellipses represent 95% confidence ellipses around
the group means.

104

Suitability of new and mature wildflower restorations as nesting sites for bees
I captured a total of 159 bees in 89 traps in the three newly-established and three
mature wildflower plantings and their paired unrestored controls, out of a total of 720
trapping events over three sample rounds in 2014 (Table 3.13). The majority of the
captured bees were subsocial members of the genus Lasioglossum (Hymenoptera:
Halictidae), with several traps capturing large numbers of the same species, presumably
from a single nest (Table 3.13). Two stem-nesting species (Ceratina dupla Say and
Hylaeus affinis (Smith)) were also abundant in the traps. One specimen of the Penstemon
oligolege Osmia distincta Cresson was captured in an old field control site; the nesting
substrate of this species has never been described (Tepedino et al., 2006).
The abundance of captured soil-nesting bees differed among habitat types (χ2 =
44.27, df = 3, p < 0.001), with significantly more bees trapped in mature wildflower
restorations than in new restorations or controls paired with new restorations (Figure 3.8).
Mature restorations averaged more bees per trap than their nearby control sites (Figure
3.8), but this difference was not significant (p = 0.11).
Bee species richness differed by habitat (χ2 = 8.64, df = 3, p = 0.03), with
significantly higher estimated richness in mature wildflower restorations and their
controls than in the unrestored controls paired with new restorations (Figure 3.9). Species
richness did not differ between the new and mature restorations or between mature
restorations and their paired controls. Simpson’s (1/D) diversity of bees in emergence
traps did not differ by habitat (χ2 = 3.76, df = 3, p = 0.29).

105

Table 3.13. Bee species collected in emergence traps from three newly-established and
three mature wildflower restorations and paired unrestored controls in 2014. No. traps
represents the number of individual traps (and therefore the minimum number of possible
nests) from which these species were collected.
Location captured
Total no. No.
New
New
Mature Mature captured traps
Control Restoration Control Restoration

Family
Species
Andrenidae
Andrena imitatrix
Perdita octamaculata
Apidae
Bombus impatiens
Ceratina dupla*
Melissodes agilis
Colletidae
Colletes nudus
Hylaeus affinis*
Halictidae
Agapostemon sericeus
Augochlorella aurata
Augochloropsis metallica
Halictus confusus
Halictus rubicundus
Lasioglossum anomalum
Lasioglossum cinctipes
Lasioglossum ellisiae
Lasioglossum foveolatum
Lasioglossum hitchensi
Lasioglossum illinoensis
Lasioglossum imitatum
Lasioglossum leucocomum
Lasioglossum leucozonium
Lasioglossum oceanicum
Lasioglossum paraforbesii
Lasioglossum pectorale
Lasioglossum perpunctatum
Lasioglossum pilosum
Megachilidae
Osmia distincta**
Total:

0
0

0
0

1
0

0
1

1
1

1
1

0
0
0

0
0
1

0
2
0

1
5
0

1
7
1

1
3
1

0
0

0
0

0
6

1
21

1
27

1
22

0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
3
0
0
4
0
5
0
0
2
3
0
0
0
0
2

2
4
0
2
0
2
1
0
1
7
0
1
0
9
1
0
2
0
0

0
5
3
0
0
8
0
0
0
6
4
0
33
0
0
3
3
1
0

2
11
3
2
3
10
1
4
1
18
4
1
35
12
1
3
5
1
2

1
8
1
2
2
6
1
1
1
8
1
1
8
7
1
1
5
1
2

0

0

1

0

1

1

2

20

42

95

159

89

* Stem-nesting bees
** Nesting substrate unknown (Tepedino et al., 2006)

106

Figure 3.8. Mean number of bees collected per emergence trap in newly-established
wildflower plantings, mature wildflower plantings, and paired unrestored controls.
Locations with the same letter are not significantly different (posthoc Dunn’s pairwise
comparisons with Bonferroni correction, α=0.05).

107

Figure 3.9. Boxplot of Chao1 estimated species richness of bees collected in emergence
traps in in newly-established wildflower plantings, mature wildflower plantings, and
paired unrestored controls. Locations with the same letter are not significantly different
(posthoc Dunn’s pairwise comparisons with Bonferroni correction, α=0.05).

Table 3.14. Microhabitat characteristics measured in emergence trap quadrats at three
new wildflower restorations, three mature wildflower restorations, and paired unrestored
control sites (n = 720 traps). Values within a column followed by the same letter are not
significantly different (posthoc Dunn’s pairwise comparisons with Bonferroni correction,
α=0.05).
Leaf
Bare ground Vegetation Wood/rocks
litter/thatch
(%) ± SEM (%) ± SEM (%) ± SEM
Habitat
(%) ± SEM
Control (new)
1.6 ± 0.6 b 95.3 ± 0.8 a
0.6 ± 0.2 b
57.8 ± 2.7 ab
New restoration
24.4 ± 2.2 a 72.2 ± 2.3 b
1.1 ± 0.2 ab
11.0 ± 1.2 c
Control (mature)
6.1 ± 1.6 b 82.7 ± 1.8 b
2.2 ± 0.3 a
47.9 ± 2.4 b
Mature restoration
1.7 ± 0.5 b 85.2 ± 1.3 b
0.6 ± 0.1 b
63.6 ± 2.8 a
"

108

Table 3.15. Spearman’s rank correlation matrix of microhabitat characteristics measured
in emergence trap quadrats at new wildflower restorations, mature wildflower
restorations, and paired unrestored control sites (n = 720 traps).
Bare ground Vegetation Wood and Leaf litter or
Slope
(%)
(%)
rocks (%)
thatch (%)
Bare ground (%)
1
Vegetation (%)
-0.54***
1
Wood and rocks (%)
0.11**
-0.17***
1
***
*
Leaf litter/thatch (%)
-0.56
0.08
0.01
1
**
***
Slope
-0.11
0.20
-0.01
0.07
1
*significant at p = 0.05, **p = 0.01, ***p = 0.001
"

Microhabitat characteristics
The microhabitat characteristics were highly intercorrelated (Table 3.15), and all of
the percent cover categories differed significantly by habitat (Table 3.14). New
restorations had significantly more bare ground than other habitats. Mature restorations
had more leaf litter/thatch than other habitats, though not significantly more than the
control habitats paired with new restorations. Control habitats paired with mature
restorations had significantly more wood and/or rocks than mature restorations or new
controls (wood and rock categories were not separated). Control habitats paired with new
restorations, which were primarily grassy field margins adjacent to blueberry fields, had
significantly more living vegetation than any other habitat type.
The microhabitat variable with the highest factor loading on PC2 was percent wood
and rocks, which loaded positively on the axis (Table 3.16). Percent leaf litter and thatch
also had a weak positive loading on the axis, but no factor besides wood and rocks had a
loading over 0.4. On PC3, the largest factor loading was a negative loading for slope, but
percent leaf litter/thatch also had a strong positive loading (Table 3.16).

109

Backward stepwise elimination led to the selection of a logit model with the second
and third principal components as predictor variables (AIC = 500.28). Individual models
with PC2 and PC3 had delta AIC values over 2 (AIC = 502.43 and 502.67, respectively),
indicating that the more parsimonious models were not viable alternatives to the
PC2+PC3 model. A likelihood ratio test indicated that the model fit for the PC2+PC3
model was significantly better than the null model (χ2 = 7.46, df = 2, p = 0.02). Bee
presence was positively correlated with both model terms (Table 3.17).

Table 3.16. Factor loadings on the first four principal components for microhabitat
characteristics in emergence trap quadrats in new wildflower restorations, mature
wildflower restorations, and their paired unrestored controls in 2014 (n=720 traps).
Factor loadings greater than 0.4 are in bold.
Microhabitat
PC1
PC2
PC3
PC4
characteristic
Bare ground (%)
0.66
0.03
-0.13
-0.16
Vegetation (%)
-0.62
-0.16
-0.10
0.41
Wood and rocks (%)
0.06
0.90
0.04
0.43
Leaf litter/thatch (%)
-0.37
0.35
0.42
-0.71
Slope
-0.20
0.22
-0.89
-0.34
"

Table 3.17. Logistic regression model assessing the second and third principal
component axes as predictors of the presence of bees in emergence traps.
Odds ratio (95%
β ± SE
Wald χ 2 df
Model term
confidence interval)
PC2
1.22 (1.00-1.46)
298.5
1
0.19 ± 0.10
PC3
1.28 (1.00-1.67)
298.4
1
0.25 ± 0.13

110

p
<0.001
<0.001

Bee communities collected in nets vs. emergence traps
NMDS ordination of trapped and netted bees indicated that the bee communities
collected by the two methods had little overlap (Figure 3.10). The bee communities
differed significantly between the two collection methods (perMANOVA; F = 2.65, df =
1, 22, p = 0.02). Twenty species contributed to the first 70% of the dissimilarity between
the two methods (Table 3.18), including a mix of soil-nesting and stem-nesting species.
The top contributors included two stem-nesting bees, Ceratina calcarata Robertson and
Hylaeus affinis (Smith), and two halictine soil-nesting bees, Lasioglossum hitchensi
Gibbs and L. leucocomum (Lovell).

Figure 3.10. Non-metric multidimensional scaling of the community of bees captured in
emergence traps (etrap, black circles) and netted on open flowers (net, gray triangles) in
mature wildflower restorations, newly-established restorations, and paired unrestored
habitats for each restoration in SW Michigan. Ellipses represent 95% confidence ellipses
around the group means.
111

Table 3.18. Bee species most responsible for differentiating bee communities by
collection method (SIMPER analysis).
Average
Total Total
Cumulative
Nesting
contribution no. in no. in
Species
contribution
substrate
traps
nets
to contrast
± SD (%)
0
23
Stem
Ceratina calcarata
7.0 ± 9.8
8.1%
18
5
Soil
Lasioglossum hitchensi
5.7 ± 8.9
14.8%
35
9
Soil
Lasioglossum leucocomum
5.6 ± 10.9
21.4%
27
4
Stem
Hylaeus affinis
4.3 ± 6.4
26.4%
11
11
Soil
Augochlorella aurata
3.8 ± 5.5
30.9%
1
13
Soil
Bombus impatiens
3.3 ± 4.5
34.7%
0
18
Stem
Hylaeus modestus
3.1 ± 8.7
38.3%
0
13
Stem
Ceratina mikmaqi
2.9 ± 7.4
41.7%
12
1
Soil
Lasioglossum leucozonium
2.9 ± 6.7
45.1%
7
2
Stem
Ceratina dupla
2.7 ± 5.7
48.3%
0
10
Soil
Bombus griseocollis
2.5 ± 3.8
51.2%
0
9
Soil
Halictus ligatus
2.3 ± 3.9
53.9%
0
9
Soil
Andrena wilkella
2.3 ± 3.0
56.5%
5
7
Soil
Lasioglossum pectorale
2.2 ± 4.3
59.1%
10
2
Soil
Lasioglossum anomalum
2.1 ± 4.1
61.5%
0
12
Soil
Lasioglossum lineatulum
2.1 ± 7.2
64.0%
3
1
Soil
Halictus rubicundus
1.7 ± 6.6
66.0%
1
7
Soil
Lasioglossum imitatum
1.5 ± 3.7
67.8%
0
3
Stem
Ceratina strenua
1.5 ± 4.9
69.5%
0
5
Soil
1.4 ± 2.7
Bombus bimaculatus
71.2%

Discussion
Using emergence traps in blueberry farms, I found that wildflower plantings
support a greater abundance of soil-nesting bees than unrestored field perimeters or
woodland in the surrounding landscape. Blueberry fields had intermediate captures.
However, species richness and diversity of soil-nesting bees did not differ across the
sampled habitats. In addition, I found evidence to suggest that the abundance of nesting
bees increases as wildflower plantings mature, as there were significantly more bees
nesting in mature wildflower plantings than in newly-established wildflower plantings.

112

Nesting bees were also more species-rich in mature wildflower plantings than new
restorations, although not more diverse.
Previous work has shown that planting wildflowers in resource-limited landscapes
can attract abundant and diverse bee communities to forage on those patches (Pywell et
al., 2005, 2007; Carvell et al., 2007; Morandin & Kremen, 2013; Blaauw & Isaacs,
2014a, 2014b). This study complements this previous work by showing that bees
preferentially nest in these undisturbed wildflower patches, indicating that wildflower
plantings provide bees with two limiting resources in these landscapes: food and nesting
sites. Other on-farm habitats may not be as well-suited for nesting: the crop field and its
margins have high levels of disturbance, including tillage and pest management inputs,
while wooded areas have thick layers of litter that may act as a barrier to nest formation.
Bee nesting was not consistently associated with specific microhabitat variables
across studies. On blueberry farms, bare ground was a significant predictor of bee
nesting, but it did not emerge as an important predictor of nesting in the restoration age
study, perhaps because variability in the percent bare ground was low in most habitats in
this study. The availability of patches of bare ground has been consistently associated
with bee nesting and/or ground-nesting bee abundance, though the strength of the
association is variable (Potts et al., 2005; Hopwood, 2008; Exeler et al., 2009; Morandin
& Kremen, 2013; Sardiñas & Kremen, 2014). Potts et al. (2005) found that species
richness of soil-nesting bees collected via timed net sampling in post-fire habitat patches
was positively associated with the percent bare ground in those habitats (p = 0.02). The
percent bare ground was also positively associated with nesting in the Sardiñas and
Kremen (2014) emergence trap study (p = 0.09). Both of these studies, however, were

113

conducted in fire-dependent scrub habitats in Mediterranean climates, where variability
in percent exposed soil is likely to be greater than in temperate prairie restorations (see
Table 3 in Potts et al., 2005 for means of microhabitat characteristics in that
environment). When variability in exposed soil is low, as in our restoration age study, bee
choice of nest site may be based on other factors, such as proximity to floral resources.
The number of small cavities (<2 cm diameter) counted in emergence trap quadrats
was the best predictor of bee presence in traps in the blueberry habitat study (Table 3.9).
Interestingly, Sardiñas and Kremen (2014) found a significant negative relationship
between the number of cavities and nesting bee abundance. However, their study placed
traps before the first emergence of bees and therefore measured pre-existing cavities from
at least the season prior to trapping, while the cavities I measured may have been a
combination of pre-existing cavities and active bee nests. More work is needed to
determine whether bees choose to nest in or to avoid pre-existing cavities, as well as to
determine the likelihood that small, round cavities in the ground are active bee nests. If
untrained observers are able to accurately identify active bee nests by sight simply by
counting cavities of a certain size and shape, then there may be faster ways to determine
bee nesting abundance in different habitats than emergence trapping.
In the restoration age study, bee presence in emergence traps was positively
associated with percent cover of leaf litter and dried thatch as well as wood and rocks,
both of which included different types of pithy stems (e.g. Rubus spp. canes were counted
as wood, while dried stems of various tall grasses were counted as thatch). We captured
several species of stem-nesting bees in emergence traps, which may explain the
association with these two cover categories (Table 3.13). Alternatively, the association

114

may be an artifact of the correlation between mature wildflower restorations and higher
levels of thatch (Table 3.14); again, it is possible that nesting bees may have been
attracted to floral resources rather than specific microhabitat characteristics.
Emergence trapping and hand netting of bees collected distinct subsets of the total
bee communities utilizing different habitat types for nesting or foraging. Differences in
collections of closely related species in nets or traps may reveal subtle niche partitioning.
For example, while all of the morphologically and genetically similar members of the
Ceratina genus nest in pithy stems, the species exhibit subtle preferences for nesting in
different plant species with exposed pith (Vickruck et al., 2011; Vickruck & Richards,
2012). We captured several Ceratina dupla in emergence traps, but no C. calcarata, C.
mikmaqi, or C. strenua, all of which were captured in nets at the same sites (Table 3.18).
Ceratina dupla may be using a different type of pithy substrate from the rest of the genus.
Similarly, two closely related species in the stem-nesting genus Hylaeus were associated
with different collection methods: Hylaeus affinis was fairly abundant in emergence
traps, with a few netted specimens, while H. modestus was captured only in nets. These
bees have long been considered nearly morphologically indistinct (Sheffield et al., 2009;
Grundel et al., 2011), and the challenge of reliable morphological delineation of the two
species means little is known about possible differences in nesting biology. The
differences in collection methods suggest there may be as-yet unknown niche partitioning
based on nesting substrate. Alternatively, these differences in collection may simply be a
sampling artifact; the spatial variability in nest sites and relatively low captures of
emerging bees mean that it’s likely that there were nesting species that were missed.

115

In addition, emergence traps captured several species of cleptoparasites in the
Nomada and Sphecodes genera in blueberry farm habitats. The apid genus Nomada
consists of cross-family cleptoparasites of andrenid bees, while the halictid genus
Sphecodes are a mixture of generalist and specialist nest parasites with hosts in various
bee families (Bogusch et al., 2006; Habermannová et al., 2013). Host associations and
their degree of specialization are not known for many cleptoparasites of soil-nesting bees,
though it has been suggested that soil nests are parasitized more frequently than
aboveground nests in twigs and other cavities (Wcislo, 1987, 1996; Wcislo & Cane,
1996; Bogusch, 2003). While cleptoparasites are sometimes captured while nectaring on
flowers, they are generally more rare in floral collection records because they do not
collect pollen, but instead co-opt the pollen resources of their host (Minckley et al.,
1999). Emergence traps may allow for better quantification of the presence and overall
proportion of cleptoparasitic species in wild bee communities than hand netting or pan
trapping. Interestingly, while the abundance of soil-nesting bees was found to be higher
in wildflower plantings than other habitat types in farm landscapes, cleptoparasites –
which might be expected to track the nesting abundance of potential hosts - were spread
relatively evenly through the different farm habitats, with cleptoparasites captured in the
crop field, wooded areas, and wildflower restorations on blueberry farms. No
cleptoparasites were captured in the three mature wildflower restorations sampled in the
restoration age study. More work is needed to determine whether increasing abundance
of soil-nesting bees over time in wildflower plantings is associated with increasing
populations of cleptoparasites and associated risk of parasitism for soil-nesting bees, but

116

these initial results suggest that wild bee cleptoparasites are not concentrated in
wildflower plantings.
Emergence traps are a promising technique for developing our understanding of bee
nesting preferences and resource requirements, and for increasing the richness of bees
sampled during community studies (Sardiñas & Kremen, 2014). However, the traps
require large investments in time and resources for variable – and sometimes quite low –
returns in terms of trap captures. To maximize return on investment, it is crucial to select
the most cost- and time-effective trapping methodology. The first trapping method, for
the on-farm study conducted from 2013-14, maximized replication with our 40
emergence traps by moving the traps to new locations every 2-3 days, and minimized
logistical issues with labor inputs by setting the traps out during daylight hours. However,
this method yielded very low capture rates overall, with an average of 0.05 ± 0.01 bees
per trap, or about 1 bee per 20 traps on average, and 1 bee per 10 traps in the wildflower
plantings. We revised the methodology for the restoration age study by placing traps after
dusk and leaving them out for slightly longer (one week) prior to collection. This method
increased logistical challenges during the peak of field season, as the driving distance to
sites ranged from 125 to 193 km from campus and hand netting of bees needed to be
conducted in suitable weather conditions, typically earlier in the day. However, this
method yielded significantly higher bees per trap, with an average of 0.22 ± 0.03 bees
captured per trap across all habitats, or around 1 bee for every 4-5 traps. In the mature
wildflower plantings, we captured about 1 bee for every 2 traps using this method (0.53 ±
0.16 bees per trap). By contrast, Sardiñas and Kremen (2014) deployed 40 larger 1.2m2
emergence traps in the same locations for the duration of the flight season (May-

117

October), returning approximately every 12 days to collect emerging bees and refill the
collection jars. This method yielded 252 bees, with 85% of traps capturing at least one
bee. These methods are not directly comparable, both because the total number of
trapping hours was much higher in the California study and because the study locations
(the Mediterranean climate of California vs. the northern temperate climate of Michigan)
are likely to differ in bee species richness and abundance. However, it is worth noting
that placing traps in early spring, rather than in midsummer as in our study, maximizes
the chance of capturing an emerging bee at some point during the season.
In choosing a method for emergence trap placement, it is also important to consider
the unit of measurement needed to answer a particular research question. The method
used here, which aimed to maximize replication with a limited number of traps, provided
an estimate of abundance of nesting bees at each site for the short windows of time
represented by each sample round. Because I did not leave the traps in one place for the
entire growing season, and therefore could not capture all of the bees that might have
been present in one 60cm2 area across multiple flight seasons, this method does not
provide an estimate of nesting density. The density of nesting bees might have important
consequences for the magnitude of bees’ effects on belowground processes, such as soil
bioturbation. Cane (2003) suggests that in the absence of a large aggregation of soilnesting bees, such as the managed alkali bee, soil-nesting bees are generally minor agents
of biogeomorphology and bioturbation. However, Colloff et al. (2010) determined that
the soil macropores generated by ground-nesting bees and other soil macroinvertebrates
are important for soil hydrological function. Density measurements of soil-nesting bees

118

would therefore be useful in determining bees’ overall contributions to soil physical and
chemical properties in different environments.
Developing effective conservation strategies for wild bees in agricultural
landscapes begins with sound information on bee resource requirements, including
knowledge of which resources may be limiting in different environments, and how bees
respond to different management approaches. Floral resources are widely regarded as the
primary limiting resource for bees in human-altered landscapes, and studies have nearly
universally shown that the installation of floral resources increases the diversity and
abundance of wild bees in these landscapes (Carvell et al., 2007; Carvalheiro et al., 2012;
Morandin & Kremen, 2013; Blaauw & Isaacs, 2014b; Jönsson et al., 2015; Wood et al.,
2015). Much less is known about whether nesting resources are a limiting factor for bee
communities, though it has been suggested for a variety of nesting guilds in different
environments (Potts & Willmer, 1997; Kremen et al., 2002; Williams et al., 2010;
Murray et al., 2012; Xie et al., 2013).
Management approaches for wild bee nesting differ by nesting guild. Widely
available artificial nesting materials for cavity and tunnel-nesting bees have been shown
to attract diverse and abundant nesting assemblages (Medler, 1966; Tscharntke et al.,
1998b; Bosch & Kemp, 2001; Sheffield et al., 2008; Junqueira et al., 2012), but there are
few proven methods for enhancing nesting by soil-dwelling bees. The creation or
maintenance of patches of bare ground, particularly in areas receiving morning or midday
sun exposure, has been widely suggested as a possible management technique to attract
soil-nesting bees (Shepherd et al., 2003; Gregory & Wright, 2005; Wild Farm Alliance,
2007; Hopwood, 2008; Mader et al., 2010; Shepherd & Vaughan, 2011; Shepherd, 2012);

119

however, there remains little empirical evidence that these practices work. This study
provides evidence from farm landscapes that wildflower restorations increase nesting by
soil-nesting bees relative to surrounding habitats, and guidance on a method that can be
used to assess the response of bees to the availability of nesting resources.

120

CHAPTER 4:

CONCLUSIONS AND FUTURE DIRECTIONS

121

In this thesis, I have explored how the distribution of resources and stressors
affect wild bee abundance, richness, and community composition in the highbush
blueberry agroecosystem. The data from two years of field sampling at farms in
southwest Michigan that were sampled almost a decade before, and at sites where habitat
enhancements have been established, revealed that wild bees are affected by various
factors that all interact at the farm level to influence these insect communities.
Long term monitoring of bee populations is rare, especially in the United States,
but this is necessary to determine how temporal or spatial changes in bee habitat quality
or quantity will affect these insects (Williams et al., 2001; Westphal et al., 2008; Lebuhn
et al., 2013). By revisiting previously-sampled farms where pesticide use has increased
over recent years, I found that the abundance, richness, diversity, and community
composition of wild bees continue to exhibit strong negative responses to insecticide
program risk (Tuell & Isaacs, 2010). Solitary bee species, including several important
Vaccinium pollinators such as Andrena carolina, show slightly greater sensitivity to
insecticide risk than social species, but the margin of difference is small.
While these results indicate a need for concern about the possible long-term
impacts on bees of pest management in fruit crops, extensive and continuous long-term
monitoring, with additional sites in unmanaged natural areas, is needed to determine the
degree and scope of any possible wild bee declines in this system (Williams et al., 2001;
Droege et al., 2003; Lebuhn et al., 2013). Using the results from 11 multiyear studies of
wild bee populations for estimates of average variability, Lebuhn et al. (2013) estimated
that a national monitoring program with 200–250 carefully selected sampling locations,
each sampled twice annually over 5 years, would provide sufficient power to detect small

122

(2–5%) annual declines in the number of species and in the total abundance of wild bees.
Other estimates have called for more frequent sampling through the season for accurate
detection of changes in regional bee communities, up to every 7-10 days for faunal
surveys (Banaszak et al., 2014). Pan traps have been nearly universally cited as the most
efficient, cost-effective, and relatively unbiased sampling method for standardized
monitoring programs (Cane et al., 2001; Droege et al., 2003, 2010; Westphal et al., 2008;
Lebuhn et al., 2013), with some caveats (Wilson et al., 2008; Baum & Wallen, 2011). An
ongoing debate remains about whether repeated lethal sampling of bees causes declines
in wild pollinators in monitored areas (Lebuhn et al., 2015; Tepedino et al., 2015), but a
recent multiyear study suggests that bee populations may be robust to this level of
disturbance (Gezon et al., 2015). Long-term monitoring across space and time, not just
within the small window of blueberry bloom, is needed for accurate estimation of the
status of wild bees in this system and beyond (Winfree, 2010; Lebuhn et al., 2013; Gezon
et al., 2015).
For blueberries in particular, I would recommend that long term monitoring to
determine changes in bee communities be established on a set of blueberry farms
stratified by insecticide use (e.g. unmanaged, low-intensity management, and highintensity management), with similar landscape composition at the local level. This
landscape caveat may be challenging to fulfill, as most unmanaged blueberry fields in
southwest Michigan that I am aware of are not surrounded by highly-managed blueberry,
unlike the managed fields included in this study. In addition to the unmanaged blueberry
sites, which provide a control habitat as similar as possible to managed blueberry, several
monitoring sites in natural areas as far removed from intensive agricultural fields as

123

possible would be helpful for distinguishing the effects of insecticide use and landscape.
Ideally, monitoring would be conducted at multiple time points through the season, with
about one sample per month from May-September at each site to maximize coverage of
different species’ flight periods (Banaszak et al., 2014).
Additional work is needed to determine how life history traits shape bee responses
to pesticide exposure, and conversely how long-term exposure to pesticides may affect
bee community structure and functional diversity (Brittain & Potts, 2011). My study
found no effect of pest management on the functional trait diversity of bee communities
in highbush blueberry fields. In the long term, however, continued differences in
sensitivity of social and solitary species to insecticide risk could lead to non-random
changes in community structure with consequences for wild plant and crop pollination
(Larsen et al., 2005; Brittain & Potts, 2011). Long-term monitoring of a set of sites using
consistent trapping methodology would allow for detection of these types of communitylevel changes.
The second part of my research, which involved the use of emergence traps to
quantify the abundance of soil-nesting bees nesting in different habitats, adds an
important piece to our understanding of wildflower restorations as a conservation strategy
for wild bees in farm landscapes. Previous work has shown that the addition of floral
resources into landscapes with scarce or fragmented floral resource patches increases the
abundance and richness of wild bees in those landscapes (Marshall et al., 2006; Carvell et
al., 2007; Pywell et al., 2007; Morandin & Kremen, 2013; Blaauw & Isaacs, 2014;
Jönsson et al., 2015; Williams et al., 2015). Morandin & Kremen (2013) suggested that
the ability of floral restorations to export pollinators to nearby crops could be the result of

124

consistent, undisturbed nesting sites as well as the availability of rich floral resources.
This study complements recent molecular surveys by Wood et al. (2015) in providing
empirical evidence that these types of restorations increase nesting by soil-nesting bees,
thereby providing dual benefits to pollinators in restored areas. Most pollinators of
blueberry - aside from managed honey bees - nest in the soil, so wildflower restorations
are a suitable choice for growers looking to support populations of blueberry-pollinating
bees on farms.
Wildflower restorations located an appropriate distance away from crop fields
may also act as refugia for wild bees in intensively-managed farm landscapes, attracting
bees to floral resources outside of treated crop areas and reducing the possibility of
pesticide exposure. Growers must weigh the benefits of placing wildflower plantings
close to crop fields to achieve pollination and biocontrol benefits (Blaauw & Isaacs,
2012, 2014) with the possibility that these areas will receive spray drift later in the
season, possibly creating a sink for pollinators nesting and foraging in restorations close
to crop fields (Davis et al., 1991; Otto et al., 2009). Several authors have suggested
buffer zones for pesticide applications near hedgerows to protect butterflies and other
organisms from the lethal and sublethal effects of insecticides, as well as to sustain the
flowering plants that might be killed by herbicide use close to hedgerows (Cuthbertson &
Jepson, 1988; Davis et al., 1991; Longley & Sotherton, 1997; Dover & Sparks, 2000).
Estimates of appropriate buffer distances from pesticide applications for honey bee hives
range from 5 to 40 meters depending on spray physical properties, formulations, and
application conditions (Davis & Williams, 1990). Similar recommendations may be
needed for the optimal placement of conservation plantings, with careful consideration of

125

the typical application methods (e.g. aerial or ground applications) and chemistries used
for pest and disease management and their potential for drift (Barret et al., 1981; Longley
et al., 1997).
My work with emergence traps provides support for the assertion of Sardiñas &
Kremen (2014) that this methodology can provide useful insights into bee nesting
requirements as well as the total community composition of different sites beyond what’s
typically captured in pan traps or nets. However, the high cost of currently available
emergence traps means that they require careful consideration of different methodologies
to receive the best return on investment. I would advise future researchers interested in
using these traps to consider the appropriate method to use to answer their specific
research questions, bearing in mind the tradeoffs of using different trapping approaches.
The method that I used to maximize replication with a limited number of traps is timeintensive and logistically difficult to integrate into a fieldwork program involving midday
bee collections. By contrast, the method employed by Sardiñas & Kremen (2014)
requires less intensive labor inputs and provides a useful estimate of season-long nest
density, but is limited in the area it can cover by the high cost of traps.
Bearing these considerations in mind, I believe the logical next step is to
determine season-long nest density of bees in different habitat types. I began sampling
bees using emergence traps in mid-June following blueberry bloom, when floral
resources were most abundant in wildflower plantings. It is possible that sampling in the
same habitats in early spring would discover higher numbers of bees nesting in forests or
forest margins, close to early-blooming resources such as flowering trees and shrubs.
This early-season sampling would also be helpful in determining the specific habitat use

126

and requirements of blueberry-pollinating bees. Mandelik et al. (2012) found that
different habitat types provide complementary floral resources for bees at different times
of the year, with concurrent shifts in bee community distribution in the landscape over
the course of the growing season; the same pattern may be true for the use of different
habitats for nesting. In addition, more accurate estimates of nest density in different
habitats will allow for a) better parameterization of pollinator abundance models based on
land cover types (e.g. Lonsdorf et al., 2011), and b) better estimation of the effects of
soil-nesting bees on soil physical and chemical properties, which have been mostly
ignored in studies of the effects of belowground organisms on soil properties (Anderson,
1994; Lavelle, 1996). For example, nest construction and food accumulation by
subterranean ants have been found to affect the vertical distribution of nutrients and
physical properties of soil relative to surrounding soil environments (Dostál et al., 2005).
These types of belowground changes can lead to changes in overall soil fertility and the
diversity and biomass production of aboveground plant communities (Hooper et al.,
2000; Wardle et al., 2004). Understanding the effects of soil-nesting bees on soil physical
and chemical properties would allow for better valuation of their total contributions to
ecosystem function.
In general, my research indicates that wild bee communities are shaped by
complex interactions of resources and stressors. Functional traits may be key to
understanding differences in species’ responses to these environmental factors, as well as
understanding how communities may change over time with unidirectional stressors like
high levels of pesticide exposure. Further research into wild bee environmental responses
should explicitly consider species-specific ecology and functional traits. While there has

127

been great interest at the national level in pollinator conservation over the last decade,
with pollinator-specific language incorporated into many different funding opportunities
from the US Department of Agriculture, we still know little about the basic ecology of
many bee species. In particular, a greater emphasis is needed on research into the
subterranean ecology of bees, so that information on nesting requirements can be
incorporated into conservation strategies for soil-nesting bees.
For blueberry growers, this research suggests that growers interested in
conserving wild bees for pollination need to employ pest management practices that
minimize risk to non-target organisms. These practices include applying sprays after
dusk, minimizing spray drift, using integrated pest management approaches to minimize
the number of applications for target pests, and choosing insecticides with lower toxicity
to bees whenever possible. In addition, growers can provide nesting and floral resources
to support wild bee populations on farms through the installation of perennial wildflower
restorations in areas protected from pesticide drift.

128

APPENDICES

129

APPENDIX A:

GPS COORDINATES FOR FIELD SITES

130

Table A. GPS coordinates for the 15 blueberry fields in which bees were sampled during
bloom by J.K. Tuell between 2004-6 and by EM in 2013 and 2014.
Site
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

N
42˚16.075
42˚20.869
42˚21.885
42˚24.749
42˚32.148
42˚34.917
42˚37.589
42˚41.677
42˚43.631
42˚49.122
42˚50.604
42˚52.051
42˚52.905
42˚57.208
42˚59.413

W
-086˚13.997
-086˚13.355
-086˚17.217
-086˚06.378
-086˚12.896
-086˚09.107
-086˚10.117
-086˚08.920
-086˚06.638
-086˚10.236
-086˚09.902
-086˚07.578
-086˚09.369
-086˚06.568
-086˚09.451

Table B. GPS coordinates for the four blueberry farms sampled using emergence traps in
2013 and 2014. Coordinates are for the wildflower restoration at each site.
Site
1
2
3
4

N
42°12.919
42°15.790
42°32.330
42°26.458

W
-086°20.338
-086°14.004
-086˚17.217
-086°05.249

Table C. GPS coordinates for the new and mature wildflower restorations sampled using
emergence traps in 2014.
Site
1
2
3
4
5
6

Type
New
New
New
Mature
Mature
Mature

N
42°22.940
42°17.920
42°58.103
42°20.931
42°21.220
43°01.041

131

W
-085°54.880
-085°50.257
-086°08.540
-085°44.345
-085°35.288
-085°50.286

APPENDIX B:

COLLECTED WILD BEE SPECIES
& ASSOCIATED FUNCTIONAL TRAITS

132

Table D. Wild bee species captured in pan traps and nets at 15 blueberry fields during bloom by J.K. Tuell between 2004-2006 and
EM in 2013 and 2014, and their associated functional traits. Trait information was taken from Wolf & Ascher (2008), Bartomeus et al.
(2013), and Gibbs et al. (2011). Key to nest location: “Soil (R)” = species that rent rather than excavate soil nests, and “Clepto” =
cleptoparasitic species. Key to sociality, “E” = eusocial, “F” = facultatively social, and “S” = solitary. Key to lecty (diet specialization):
“P” = polylectic and “O” = oligolectic. Body size was categorized from inter-tegular distance, where available, or based on educated
guesses (Bartomeus et al., 2013). Key to body size: “S” = small, “M” = medium, and “L” = large (Table 2.2). Service-providing bees
are those that have been recorded collecting pollen from Vaccinium corymbosum L. and close relatives (see reference column); a “no”
in the service-providing column means that the species has not been recorded visiting Vaccinium flowers, or is a cleptoparasite or an
oligolege on other types of plants.
Total no. captured
Functional traits
Nest
Body
Service
2004 2005 2006 2013 2014
Sociality Lecty
Species
location
size
providing Ref
3
Agapostemon sericeus
1
3
4
4
0
Soil
F
P
M
Yes
4
Agapostemon splendens
2
1
0
2
2
Soil
F
P
M
Yes
2
Agapostemon texanus
0
2
1
4
0
Soil
F
P
M
Yes
5
Agapostemon virescens
3
5
3
2
3
Soil
F
P
M
Yes
Andrena algida
1
1
0
0
0
Soil
S
P
M
No
2
Andrena alleghaniensis
8
21
18
0
1
Soil
S
P
M
Yes
Andrena andrenoides
0
1
0
0
0
Soil
S
P
M
No
Andrena arabis
0
0
1
0
0
Soil
S
O
M
No
Andrena barbilabris
0
8
2
1
0
Soil
S
P
M
No
1
Andrena carlini
57
55
17
0
7
Soil
S
P
L
Yes
1,2,3,10
Andrena carolina
111
303 113
8
16
Soil
S
O
M
Yes
6
Andrena ceanothi
4
2
9
2
0
Soil
S
P
L
Yes
3
Andrena cf. morrisonella
1
0
0
0
0
Soil
S
P
M
Yes
2
Andrena cf. w-scripta
0
0
0
0
1
Soil
S
P
M
Yes
6
Andrena clarkella
0
1
0
0
0
Soil
S
O
M
Yes
Andrena commoda
2
2
0
0
0
Soil
S
P
M
No
Andrena confederata
1
0
0
0
0
Soil
S
P
M
No

133

Table D (cont’d)
Andrena crataegi
Andrena cressonii
Andrena dunningi
Andrena erigeniae
Andrena erythrogaster
Andrena erythronii
Andrena forbesii
Andrena hilaris
Andrena hippotes
Andrena imitatrix
Andrena imitatrix or
morrisonella
Andrena integra
Andrena mandibularis
Andrena mariae
Andrena milwaukeensis
Andrena miranda
Andrena miserabilis
Andrena morrisonella
Andrena nasonii
Andrena neonana
Andrena nigrae
Andrena nigrihirta
Andrena nivalis
Andrena nuda
Andrena perplexa
Andrena persimulata

17
9
1
0
0
0
5
3
7
31

10
23
0
2
1
0
0
0
5
26

16
13
0
0
0
1
7
0
4
21

0
3
1
0
0
0
1
0
0
1

3
9
2
1
0
0
3
0
1
6

Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil

S
S
S
S
S
S
S
S
S
S

P
P
P
O
O
O
P
P
P
P

M
M
M
S
M
M
M
M
M
M

Yes
Yes
Yes
No
No
No
Yes
No
No
Yes

8
2
0
1
1
0
9
3
7
1
3
0
1
15
15
0

27
0
2
0
0
0
45
4
14
0
0
1
0
4
6
0

18
0
0
0
0
0
43
3
11
0
11
7
0
7
3
1

0
0
0
0
0
1
3
0
4
0
0
0
0
0
3
0

0
0
0
0
0
0
4
0
9
0
0
0
0
1
2
0

Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil

S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S

P
O
P
O
P
P
P
P
P
P
O
P
P
P
P
O

M
M
M
M
M
M
M
M
M
M
M
M
L
M
M
M

Yes
No
Yes
No
Yes
Yes
Yes
Yes
Yes
No
No
No
Yes
No
Yes
No

134

2
1
2

2,7

1,2
1,2,3
2
2
6
2, 3
3
3

1,2,10
3

Table D (cont’d)
Andrena pruni
Andrena rehni
Andrena robertsonii
Andrena rugosa
Andrena salictaria
Andrena sp.
Andrena spiraeana
Andrena vicina
Andrena wilkella
Augochlora pura
Augochlorella aurata
Augochloropsis metallica
Bombus bimaculatus
Bombus citrinus
Bombus fervidus
Bombus griseocollis
Bombus impatiens
Bombus perplexus
Bombus vagans
Ceratina calcarata
Ceratina dupla
Ceratina mikmaqi
Ceratina strenua
Colletes inaequalis
Colletes thoracicus
Colletes validus
Eucera atriventris

2
2
1
10
1
2
0
53
0
41
68
1
7
15
0
3
1
3
0
43
1
16
9
0
24
0
1

1
0
1
4
5
10
1
49
0
13
32
0
0
0
3
0
2
0
0
59
0
15
6
15
4
2
0

1
0
2
6
0
2
0
26
0
11
72
0
2
1
3
4
0
0
0
44
0
6
0
3
0
0
0

0
0
0
1
0
1
0
4
1
15
40
0
1
1
7
0
1
0
0
15
0
7
3
1
1
0
0

0
0
0
3
0
8
0
2
0
17
15
0
10
0
9
4
53
0
9
33
0
2
2
5
3
0
1

135

Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Wood
Soil
Soil
Soil (R)
Clepto
Soil (R)
Soil (R)
Soil (R)
Soil (R)
Soil (R)
Stem
Stem
Stem
Stem
Soil
Soil
Soil
Soil

S
S
S
S
S
S
S
S
S
S
E
S
E
C
E
E
E
E
E
F
F
F
F
S
S
S
S

P
P
P
P
O
P
P
P
P
P
P
P
P
C
P
P
P
P
P
P
P
P
P
P
P
P
P

M
M
M
M
M
M
M
M
M
M
M
M
L
L
L
L
L
L
L
S
S
S
S
L
L
L
L

Yes
No
No
Yes
Yes
Unknown
No
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No

7

2
2

1
1,3
1
3
1,7
2
7
11
10
2
2
7
2
7!
7
3
3
3,10

Table D (cont’d)
Eucera hamata
Halictus confusus
Halictus ligatus
Halictus parallelus
Halictus rubicundus
Hoplitis pilosifrons
Hoplitis producta
Hoplitis spoliata
Hylaeus affinis
Hylaeus illinoisensis
Hylaeus mesillae
Hylaeus sp.
Lasioglossum abanci
Lasioglossum acuminatum
Lasioglossum admirandum
Lasioglossum anomalum
Lasioglossum atwoodi
Lasioglossum bruneri
Lasioglossum cattellae
Lasioglossum coeruleum
Lasioglossum coriaceum
Lasioglossum cressonii
Lasioglossum ellisiae
Lasioglossum ephialtum
Lasioglossum floridanum
Lasioglossum foxii
Lasioglossum fuscipenne

0
2
17
2
3
1
1
0
1
1
0
0
0
5
11
1
0
3
0
5
60
80
6
0
3
2
4

1
27
20
1
7
1
2
1
2
1
1
2
0
3
3
4
0
1
1
1
17
17
5
0
4
3
1

0
4
11
2
3
0
0
0
1
2
0
0
0
4
4
2
0
0
1
3
12
17
16
0
0
3
1

5
1
2
5
1
0
2
0
0
0
0
0
0
2
2
1
1
0
0
0
42
34
3
8
0
0
0

1
7
25
10
4
2
0
0
1
0
0
0
2
7
5
25
0
0
0
3
61
39
13
1
0
1
6

136

Soil
Soil
Soil
Soil
Soil
Stem
Stem
Stem
Cavity
Cavity
Cavity
Cavity
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Wood
Soil
Wood
Soil
Soil
Soil
Soil
Soil

S
F
E
E
F
S
S
S
S
S
S
S
E
F
E
E
E
E
E
E
F
E
E
E
E
S
F

P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P

L
S
M
L
M
M
S
S
S
S
S
S
S
M
M
S
S
S
S
S
M
S
S
S
S
S
M

No
Yes
Yes
Yes
Yes
No
No
No
Yes
No
Yes
Unknown
No
No
Yes
No
No
No
No
No
Yes
Yes
No
No
No
Yes
No

2
2
5
3,8,10

1
2

2

1,3
1,7

2

Table D (cont’d)
Lasioglossum hitchensi
Lasioglossum illinoense
Lasioglossum imitatum
Lasioglossum laevissimum
Lasioglossum leucocomum
Lasioglossum leucozonium
Lasioglossum lineatulum
Lasioglossum lustrans
Lasioglossum macoupinense
Lasioglossum nelumbonis
Lasioglossum nigroviride
Lasioglossum nymphaearum
Lasioglossum oblongum
Lasioglossum obscurum
Lasioglossum oceanicum
Lasioglossum paradmirandum
Lasioglossum pectorale
Lasioglossum perpunctatum
Lasioglossum pilosum
Lasioglossum planatum
Lasioglossum quebecense
Lasioglossum sagax
Lasioglossum smilacinae
Lasioglossum subviridatum
Lasioglossum timothyi
Lasioglossum versans
Lasioglossum versatum

23
0
25
0
8
64
5
0
3
1
2
1
5
2
0
0
11
4
29
1
1
1
0
1
6
8
22

9
1
72
0
28
27
20
0
2
0
1
1
5
0
0
0
14
3
112
0
1
0
0
2
2
1
2

10
0
27
0
20
56
0
1
4
0
7
2
3
0
0
0
14
3
40
0
0
0
0
0
1
4
2

0
0
0
0
8
12
0
0
9
1
0
1
0
0
0
0
14
1
68
0
0
0
0
4
8
3
0

7
0
39
1
39
18
12
0
9
2
4
0
2
2
4
1
19
1
85
0
0
0
3
0
0
1
1

137

Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Wood
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Soil
Wood
Soil
Soil
Soil

E
E
E
E
E
S
E
S
S
E
E
E
E
E
E
E
S
E
E
E
S
E
E
E
E
E
E

P
P
P
P
P
P
P
O
P
O
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P
P

S
S
S
S
S
M
S
S
S
S
M
S
S
S
S
S
S
S
S
S
M
S
M
S
S
S
S

No
No
Yes
Yes
Yes
Yes
No
No
No
No
No
No
Yes
No
No
No
Yes
Yes
Yes
No
Yes
No
Yes
No
No
Yes
Yes

2
1
10
2

2

5
2
2
1,10
7

1
7

Table D (cont’d)
Lasioglossum vierecki
Lasioglossum viridatum
Lasioglossum zephyrum
Lasioglossum zonulum
Megachile addenda
Megachile gemula
Megachile mucida
Megachile rotundata
Nomada cressonii
Nomada denticulata
Nomada maculata
Nomada ovata
Nomada sp.
Osmia albiventris
Osmia atriventris
Osmia bucephala
Osmia caerulescens
Osmia conjuncta
Osmia cornifrons
Osmia felti
Osmia georgica
Osmia pumila
Osmia virga
Sphecodes cressonii
Sphecodes confertus
Sphecodes dichrous
Sphecodes ranunculi

1
1
0
0
1
1
0
0
0
0
0
0
33
1
0
4
0
1
0
1
0
4
0
0
2
0
2

6
1
1
0
0
0
2
0
1
2
1
1
47
0
1
2
0
2
0
0
0
4
1
0
0
0
0

0
1
0
0
0
0
0
1
0
0
0
0
13
0
1
0
0
0
1
1
0
4
0
0
0
0
0

0
1
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0

2
1
1
2
0
0
0
0
0
1
0
0
0
0
3
1
1
0
0
0
1
5
0
1
0
1
0

138

Soil
Soil
Soil
Soil
Cavity
Cavity
Cavity
Cavity
Clepto
Clepto
Clepto
Clepto
Clepto
Cavity
Cavity
Cavity
Cavity
Cavity
Cavity
Cavity
Cavity
Cavity
Cavity
Clepto
Clepto
Clepto
Clepto

S
E
E
S
S
S
S
S
C
C
C
C
C
S
S
S
S
S
S
S
S
S
S
C
C
C
C

P
P
P
P
P
P
P
P
C
C
C
C
C
P
P
P
P
P
P
P
P
P
P
C
C
C
C

S
S
S
M
M
L
L
M
M
M
M
M
M
M
M
L
M
M
L
M
M
M
M
M
S
M
M

No
Yes
Yes
Yes
No*
Yes
No
Yes
No
No
No
No
No
No
Yes
Yes
No
No
Yes
Yes
No
Yes
Yes
No
No
No
No

2
7
2
1
2,8

9
7

5
1
1
2

Table D (cont’d)
Sphecodes sp.
Xylocopa virginica

2
2

0
1

2
0

0
3

0
22

Clepto
Wood

C
F

C
P

M
L

No
Yes

1

References: 1) Cane et al., 1985; 2) Stubbs et al., 1992; 3) Tuell et al., 2009; 4) Deyrup et al., 2002; 5) Adamson, 2011; 6) Sheffield et
al., 2003; 7) MacKenzie & Eickwort, 1996; 8) Javorek et al., 2002; 9) Stubbs & Drummond, 1997; 10) Moisan-Deserres et al., 2014;
11) Benjamin et al., 2014.
! Inferred from other Ceratina references, as Ceratina mikmaqi was first described from the Ceratina dupla Say species complex in
2011 (Rehan & Sheffield, 2011).
* Megachile addenda is an excellent pollinator of cranberry (Vaccinium macrocarpon) (Cane, 1996), but has not been recorded on
Vaccinium corymbosum

139

APPENDIX C:

RECORD OF DEPOSITION OF VOUCHER SPECIMENS

140

RECORD OF DEPOSITION OF VOUCHER SPECIMENS
The specimens listed below have been deposited in the named museum as samples of
those species or other taxa used in this research. Voucher recognition labels bearing the
voucher number have been attached to pinned specimens.
Voucher Number: 2015-01
Author and title of thesis: Emily A. May, “Wild bee community responses to farm
management practices, wildflower restorations, and landscape composition.”
Museum where deposited: Albert J. Cook Arthropod Research Collection, Michigan State
University (MSU)
Table E. List of voucher specimens. Each record represents one pinned adult specimen.
Species determinations (“Det”) were made by Dr. Jason Gibbs (JG) and the author (EM).
Unique
Date
ID
Family
Genus-Species
collected Sex
Det
13-0392 Halictidae
Agapostemon sericeus
5/29/13
F
EM
13-0093 Halictidae
Agapostemon splendens
5/20/13
F
EM
13-0130 Halictidae
Agapostemon texanus
5/21/13
F
EM
13-0310 Halictidae
Agapostemon virescens
5/29/13
F
EM
14-0112 Andrenidae
Andrena alleghaniensis
5/24/14
F
JG
14-1027 Andrenidae
Andrena asteris
9/18/14
F
JG
13-0025 Andrenidae
Andrena barbilabris
5/20/13
F
JG
14-1452 Andrenidae
Andrena canadensis
9/3/14
F
JG
13-0418 Andrenidae
Andrena carlini
5/30/13
F
JG
13-0317 Andrenidae
Andrena carolina
5/29/13
F
JG
13-0292 Andrenidae
Andrena ceanothi
5/29/13
F
JG
14-0076 Andrenidae
Andrena cf. w-scripta
5/26/14
F
JG
14-0786 Andrenidae
Andrena commoda
6/27/14
F
JG
14-0006 Andrenidae
Andrena crataegi
5/21/14
M
JG
14-0508 Andrenidae
Andrena cressonii
5/24/14
F
JG
13-0342 Andrenidae
Andrena dunningi
5/30/13
F
JG
14-0282 Andrenidae
Andrena erigeniae
5/21/14
F
JG
13-0210 Andrenidae
Andrena forbesii
5/30/13
F
JG
14-0509 Andrenidae
Andrena hippotes
5/24/14
F
JG
14-1024 Andrenidae
Andrena hirticincta
9/8/14
F
JG
14-0252 Andrenidae
Andrena imitatrix
5/29/14
M
JG
13-0204 Andrenidae
Andrena miserabilis
5/30/13
F
JG

141

Table E (cont’d)
13-0079 Andrenidae
14-0559 Andrenidae
13-0194 Andrenidae
14=1010 Andrenidae
13-0001 Andrenidae
13-0097 Andrenidae
14-0799 Andrenidae
13-028
Megachilidae
13-007
Apidae
14-1034 Apidae
14-0837 Apidae
13-0356 Halictidae
13-0295 Halictidae
14-1346 Halictidae
13-0419 Apidae
13-0082 Apidae
13-0052 Apidae
13-0420 Apidae
13-0406 Apidae
14-0976 Apidae
13-0337 Apidae
14-1449 Apidae
13-0319 Apidae
13-0343 Apidae
13-0369 Colletidae
14-1454 Colletidae
14-0638 Colletidae
13-0217 Colletidae
13-057
Halictidae
14-0001 Apidae
14-0004 Apidae
14-0438 Halictidae
14-0253 Halictidae
14-0150 Halictidae
14-0599 Halictidae
14-0717 Megachilidae
14-0554 Megachilidae
13-0388 Megachilidae
14-0898 Megachilidae
14-0065 Colletidae
14-0956 Colletidae
14-0790 Colletidae

Andrena nasonii
Andrena nuda
Andrena perplexa
Andrena placata
Andrena rugosa
Andrena vicina
Andrena wilkella
Anthidium oblongatum
Anthophora bomboides
Anthophora terminalis
Apis mellifera
Augochlora pura
Augochlorella aurata
Augochloropsis metallica
Bombus bimaculatus
Bombus citrinus
Bombus fervidus
Bombus griseocollis
Bombus impatiens
Bombus vagans
Ceratina calcarata
Ceratina dupla
Ceratina mikmaqi
Ceratina strenua
Colletes inaequalis
Colletes nudus
Colletes thoracicus
Colletes thoracicus
Dieunomia heteropoda
Eucera atriventris
Eucera hamata
Halictus confusus
Halictus ligatus
Halictus parallelus
Halictus rubicundus
Heriades carinata
Hoplitis pilosifrons
Hoplitis producta
Hoplitis spoliata
Hylaeus affinis
Hylaeus annulatus
Hylaeus mesillae
142

5/21/13
5/24/14
5/30/13
8/15/14
5/17/13
5/20/13
6/24/14
6/21/13
6/19/13
9/8/14
6/30/14
5/29/13
5/29/13
6/10/14
5/30/13
5/20/13
5/21/13
5/30/13
5/21/13
8/5/14
5/29/13
8/7/14
5/29/13
5/30/13
5/29/13
7/17/14
5/26/14
5/30/13
7/17/13
5/20/14
5/20/14
5/25/14
5/29/14
5/29/14
5/24/14
6/16/14
5/25/14
5/29/13
7/22/14
5/26/14
8/5/14
6/24/14

F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
M
F
M
M
F
F
F
F
F
F
F
F
M
F
F

JG
JG
JG
JG
JG
JG
JG
JG
EM
JG
EM
EM
EM
JG
EM
EM
EM
EM
EM
EM
EM
JG
EM
EM
EM
JG
EM
EM
EM
EM
EM
EM
EM
EM
EM
JG
JG
EM
JG
JG
JG
JG

Table E (cont’d)
14-0828 Colletidae
14-0266 Halictidae
14-0289 Halictidae
13-0129 Halictidae
14-0473 Halictidae
13-0073 Halictidae
14-0249 Halictidae
14-0110 Halictidae
13-0230 Halictidae
13-0298 Halictidae
14-0407 Halictidae
13-049
Halictidae
14-1474 Halictidae
14-0342 Halictidae
14-0348 Halictidae
13-0256 Halictidae
14-0261 Halictidae
14-0321 Halictidae
14-0456 Halictidae
14-0595 Halictidae
14-0470 Halictidae
14-0873 Halictidae
14-0081 Halictidae
13-0355 Halictidae
14-0405 Halictidae
14-0275 Halictidae
14-0304 Halictidae
13-0247 Halictidae
14-0621 Halictidae
13-0314 Halictidae
14-0516 Halictidae
14-0371 Halictidae
14-0286 Halictidae
13-0593 Halictidae
14-0810 Halictidae
13-0116 Halictidae
14-1000 Halictidae
13-0349 Halictidae
14-0283 Halictidae
14-0335 Halictidae
13-0255 Halictidae
14-0388 Halictidae

Hylaeus modestus
Lasioglossum abanci
Lasioglossum acuminatum
Lasioglossum admirandum
Lasioglossum anomalum
Lasioglossum atwoodi
Lasioglossum coeruleum
Lasioglossum coriaceum
Lasioglossum cressonii
Lasioglossum ellisiae
Lasioglossum ephialtum
Lasioglossum floridanum
Lasioglossum foveolatum
Lasioglossum foxii
Lasioglossum fuscipenne
Lasioglossum hitchensi
Lasioglossum imitatum
Lasioglossum laevissimum
Lasioglossum leucocomum
Lasioglossum leucozonium
Lasioglossum lineatulum
Lasioglossum lustrans
Lasioglossum macoupinense
Lasioglossum nelumbonis
Lasioglossum nigroviride
Lasioglossum oblongum
Lasioglossum obscurum
Lasioglossum oceanicum
Lasioglossum paradmirandum
Lasioglossum pectorale
Lasioglossum perpunctatum
Lasioglossum pilosum
Lasioglossum smilacinae
Lasioglossum subviridatum
Lasioglossum tegulare
Lasioglossum timothyi
Lasioglossum trigeminum
Lasioglossum versans
Lasioglossum versatum
Lasioglossum vierecki
Lasioglossum viridatum
Lasioglossum zephyrum
143

7/10/14
5/21/14
5/21/14
5/21/13
5/24/14
5/21/13
5/29/14
5/24/14
5/30/13
5/29/13
5/24/14
7/1/13
7/17/14
5/20/14
5/20/14
5/30/13
5/21/14
5/20/14
5/24/14
5/30/14
5/24/14
7/25/14
5/29/14
5/29/13
5/24/14
5/21/14
5/20/14
5/30/13
5/30/14
5/29/13
5/24/14
5/24/14
5/21/14
7/25/13
7/17/14
5/21/13
8/20/14
5/29/13
5/21/14
5/25/14
5/30/13
5/24/14

F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F

JG
JG
JG
JG
JG
JG
JG
EM
JG
JG
JG
JG
JG
JG
EM
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
JG
EM
JG
JG

Table E (cont’d)
14-0618 Halictidae
14-1401 Megachilidae
14-0923 Megachilidae
14-0924 Megachilidae
14-0732 Megachilidae
14-0962 Megachilidae
14-0918 Megachilidae
14-0872 Megachilidae
14-0739 Megachilidae
14-1422 Apidae
14-0968 Apidae
13-058
Apidae
14-0974 Apidae
14-1045 Apidae
13-046
Apidae
14-0687 Apidae
14-0043 Apidae
13-008
Apidae
14-1466 Apidae
14-0408 Megachilidae
14-0639 Megachilidae
14-0417 Megachilidae
14-1429 Megachilidae
14-0240 Megachilidae
14-0634 Megachilidae
14-0699 Andrenidae
14-1407 Andrenidae
14-1033 Andrenidae
14-0809 Halictidae
14-0334 Halictidae
13-009
Halictidae
14-0219 Halictidae
13-021
Halictidae
13-0410 Apidae

Lasioglossum zonulum
Megachile brevis
Megachile centucularis
Megachile gemula
Megachile latimanus
Megachile mendica
Megachile petulans
Megachile pugnata
Megachile texana
Melissodes agilis
Melissodes bimaculata
Melissodes communis
Melissodes desponsa
Melissodes druriella
Melissodes subillata
Nomada articulata
Nomada denticulata
Nomada illinoensis
Nomada vegana
Osmia atriventris
Osmia bucephala
Osmia caerulescens?
Osmia distincta
Osmia georgica
Osmia pumila
Perdita halictoides
Perdita octomaculata
Pseudopanurgus nebrascensis
Sphecodes cf. coronus
Sphecodes cressonii
Sphecodes davisii
Sphecodes dichrous
Sphecodes mandicularis
Xylocopa virginica

144

5/30/14
8/18/14
7/21/14
8/15/14
6/25/14
8/5/14
7/21/14
7/30/14
6/25/14
8/14/14
8/5/14
7/17/13
8/5/14
9/9/14
7/1/13
6/16/14
5/29/14
6/19/13
7/8/14
5/20/04
5/26/14
5/20/14
6/24/14
5/29/14
5/25/14
6/16/14
8/18/14
9/8/14
7/7/14
5/25/14
6/19/13
5/20/14
7/19/13
5/21/13

F
F
F
M
F
F
M
M
M
F
F
F
M
F
F
F
F
M
M
F
F
M
F
M
F
F
F
F
M
F
F
F
F
F

JG
EM
JG
JG
JG
JG
JG
JG
JG
JG
JG
EM
JG
JG
JG
JG
JG
JG
JG
EM
EM
EM
JG
EM
EM
JG
JG
JG
JG
JG
JG
JG
JG
EM

LITERATURE CITED

145

LITERATURE CITED
Adamson, N.L. (2011) An assessment of non-Apis bees as fruit and vegetable crop
pollinators in southwest Virginia. Dissertation. Virginia Polytechnic Institute and
State University, Blacksburg, VA.
Ahrné, K., Bengtsson, J. & Elmqvist, T. (2009) Bumble bees (Bombus spp) along a
gradient of increasing urbanization. PLoS ONE, 4, e5574.
Aizen, M.A., Garibaldi, L.A., Cunningham, S.A. & Klein, A.-M. (2008a) Long-term
global trends in crop yield and production reveal no current pollination shortage
but increasing pollinator dependency. Current Biology, 18, 1572–1575.
Aizen, M.A. & Harder, L.D. (2009) The global stock of domesticated honey bees is
growing slower than agricultural demand for pollination. Current Biology, 19,
915–918.
Aizen, M.A., Morales, C.L. & Morales, J.M. (2008b) Invasive mutualists erode native
pollination webs. PLoS Biology, 6, 397–403.
Alarcón, R., Waser, N.M. & Ollerton, J. (2008) Year-to-year variation in the topology of
a plant-pollinator interaction network. Oikos, 117, 1796–1807.
Alaux, C., Brunet, J.L., Dussaubat, C., Mondet, F., Tchamitchan, S., Cousin, M., et al.
(2010a) Interactions between Nosema microspores and a neonicotinoid weaken
honeybees (Apis mellifera). Environmental Microbiology, 12, 774–782.
Alaux, C., Ducloz, F., Crauser, D. & Conte, Y. Le. (2010b) Diet effects on honeybee
immunocompetence. Biology Letters, 6, 562–565.
Albrecht, M., Schmid, B., Hautier, Y. & Muller, C.B. (2012) Diverse pollinator
communities enhance plant reproductive success. Proceedings of the Royal Society
B: Biological Sciences, 279, 4845–4852.
Allen-Wardell, G., Bernhardt, P., Bitner, R., Burquez, A., Buchmann, S., Cane, J., et al.
(1998) The potential consequences of pollinator declines on the conservation of
biodiversity and stability of food crop yields. Conservation Biology, 12, 8–17.
Anderson, J.M. (1994) Soil organisms as engineers: Microscale modulation of
macroscale processes. In Linking Species and Ecosystems (ed. by Jones, C. &
Lawton, J.). Chapman and Hall, New York, pp. 94–106.
Arena, M. & Sgolastra, F. (2014) A meta-analysis comparing the sensitivity of bees to
pesticides. Ecotoxicology, 23, 324–34.

146

Bale, J.S. & Hayward, S.A.L. (2010) Insect overwintering in a changing climate. The
Journal of Experimental Biology, 213, 980–994.
Banaszak, J., Banaszak-Cibicka, W. & Szefer, P. (2014) Guidelines on sampling intensity
of bees (Hymenoptera: Apoidea: Apiformes). Journal of Insect Conservation, 18,
651–656.
Barret, R.E., Maas, J.L., Retzer, H.J. & Adams, R.E. (1981) Comparison of spray droplet
size, pesticide deposition, and drift with ultralow-volume, low-volume, and dilute
pesticide application on apple. Plant Disease, 65, 872–875.
Bartomeus, I., Ascher, J.S., Gibbs, J., Danforth, B.N., Wagner, D.L., Hedtke, S.M., et al.
(2013) Historical changes in northeastern US bee pollinators related to shared
ecological traits. Proceedings of the National Academy of Sciences of the United
States of America, 110, 4656–4660.
Bartomeus, I., Ascher, J.S., Wagner, D., Danforth, B.N., Colla, S., Kornbluth, S., et al.
(2011) Climate-associated phenological advances in bee pollinators and beepollinated plants. Proceedings of the National Academy of Sciences of the United
States of America, 108, 20645–9.
Barton, K. (2014) MuMIn: Multi-model inference. R package version 1.10.5. Available:
http://cran.r-project.org/package=MuMIn.
Bascompte, J., Jordano, P., Melián, C.J. & Olesen, J.M. (2003) The nested assembly of
plant-animal mutualistic networks. Proceedings of the National Academy of
Sciences of the United States of America, 100, 9383–9387.
Bates, A.J., Sadler, J.P., Fairbrass, A.J., Falk, S.J., Hale, J.D. & Matthews, T.J. (2011)
Changing bee and hoverfly pollinator assemblages along an urban-rural gradient.
PLoS ONE, 6, e23459.
Batra, S.W.T. (1985) Red maple (Acer rubrum L.), an important early spring food
resource for honey bees and other insects. Journal of the Kansas Entomological
Society, 58, 169–172.
Baum, K.A. & Wallen, K.E. (2011) Potential bias in pan trapping as a function of floral
abundance. Journal of the Kansas Entomological Society, 84, 155–159.
Benton, T.G., Vickery, J.A. & Wilson, J.D. (2003) Farmland biodiversity: is habitat
heterogeneity the key? Trends in Ecology and Evolution, 18, 182–188.
Benjamin, F.E., Reilly, J.R. & Winfree, R. (2014) Pollinator body size mediates the scale
at which land use drives crop pollination services. Journal of Applied Ecology, 51,
440–449.

147

Biddinger, D.J., Robertson, J.L., Mullin, C., Frazier, J., Ashcraft, S.A., Rajotte, E.G., et
al. (2013) Comparative toxicities and synergism of apple orchard pesticides to Apis
mellifera (L.) and Osmia cornifrons (Radoszkowski). PLoS ONE, 8, e72587.
Biesmeijer, J.C., Roberts, S.P.M., Reemer, M., Ohlemuller, R., Edwards, M., Peeters, T.,
et al. (2006) Parallel declines in pollinators and insect-pollinated plants in Britain
and the Netherlands. Science, 313, 351–354.
Blaauw, B.R. & Isaacs, R. (2012) Larger wildflower plantings increase natural enemy
density, diversity, and biological control of sentinel prey, without increasing
herbivore density. Ecological Entomology, 37, 386–394.
Blaauw, B.R. & Isaacs, R. (2014a) Larger patches of diverse floral resources increase
insect pollinator density, diversity, and their pollination of native wildflowers.
Basic and Applied Ecology, 15, 701–711.
Blaauw, B.R. & Isaacs, R. (2014b) Flower plantings increase wild bee abundance and the
pollination services provided to a pollination-dependent crop. Journal of Applied
Ecology, 51, 890–898.
Blüthgen, N. & Klein, A.-M. (2011) Functional complementarity and specialisation: The
role of biodiversity in plant–pollinator interactions. Basic and Applied Ecology, 12,
282–291.
Bolda, M.P., Al, E., Goodhue, R.E. & Zalom, F.G. (2010) Spotted wing drosophila:
Potential economic impact of a newly established pest. Agricultural and Resource
Economics Update, 13, 5–8.
Borcard, D., Gillet, F. & Legendre, P. (2011) Numerical Ecology with R. Springer. 306 p.
Brewer, J.W., Dobson, R.C. & Nelson, J.W. (1969) Effects of increased pollinator levels
on production of the highbush blueberry, Vaccinium corymbosum. Journal of
Economic Entomology, 62, 815–818.
Brittain, C., Bommarco, R., Vighi, M., Barmaz, S., Settele, J. & Potts, S.G. (2010a) The
impact of an insecticide on insect flower visitation and pollination in an
agricultural landscape. Agricultural and Forest Entomology, 12, 259–266.
Brittain, C. & Potts, S.G. (2011) The potential impacts of insecticides on the life-history
traits of bees and the consequences for pollination. Basic and Applied Ecology, 12,
321–331.
Brittain, C., Vighi, M., Bommarco, R., Settele, J. & Potts, S.G. (2010) Impacts of
pesticide on pollinator species richness at different spatial scales. Basic and
Applied Ecology, 11, 106–115.

148

Brodschneider, R. & Crailsheim, K. (2010) Nutrition and health in honey bees.
Apidologie, 41, 278–294.
Brosi, B.J., Daily, G.C. & Ehrlich, P.R. (2007) Bee community shifts with landscape
context in a tropical countryside. Ecological Applications, 17, 418–430.
Buchmann, S.L. (1983) Buzz pollination in angiosperms. In Handbook of Experimental
Pollination Biology (ed. by Jones, C.E. & Little, R.J.). Van Nostrand Reinhold,
New York, NY, pp. 73–113.
Burnham, K. P., & Anderson, D. R. (2002) Model Selection and Multimodel Inference: A
Practical Information-Theoretic Approach. 2nd Edition. Springer-Verlag, New
York. 488 p.
Buschini, M.L.T. (2006) Species diversity and community structure in trap-nesting bees
in Southern Brazil. Apidologie, 37, 58–66.
Cailliez, F. (1983) The analytical solution of the additive constant problem.
Psychometrika, 48, 305–308.
Cameron, S.A., Lozier, J.D., Strange, J.P., Koch, J.B., Cordes, N., Solter, L.F., et al.
(2011) Patterns of widespread decline in North American bumble bees.
Proceedings of the National Academy of Sciences of the United States of America,
108, 662–7.
Cane, J.H. (1991) Soils of ground-nesting bees (Hymenoptera: Apoidea): texture,
moisture, cell depth, and climate. Journal of the Kansas Entomological Society,
64, 406–413.
Cane, J.H., Eickwort, G.C., Wesley, F.R. & Spielholz, J. (1985) Pollination ecology of
Vaccinium stamineum (Ericaceae: Vaccinioideae). American Journal of Botany,
72, 135–142.
Cane, J.H., Griswold, T. & Parker, F.D. (2007) Substrates and materials used for nesting
by North American Osmia bees (Hymenoptera: Apiformes: Megachilidae). Annals
of the Entomological Society of America, 100, 350–358.
Cane, J.H., Minckley, R.L. & Kervin, L. (2001) Sampling bees (Hymenoptera:
Apiformes) for pollinator community studies: pitfalls of pan trapping. Journal of
the Kansas Entomological Society, 73, 208–214.
Cane, J.H. & Neff, J.L. (2011) Predicted fates of ground-nesting bees in soil heated by
wildfire: Thermal tolerances of life stages and a survey of nesting depths.
Biological Conservation, 144, 2631–2636.

149

Cane, J.H. & Payne, J.A. (1993) Regional, annual, and seasonal variation in pollinator
guilds: Intrinsic traits of bees (Hymenoptera: Apoidea) underlie their patterns of
abundance at Vaccinium ashei (Ericaceae). Annals of the Entomological Society of
America, 85, 577–588.
Cane, J.H. & Tepedino, V.J. (2001) Causes and extent of declines among native North
American invertebrate pollinators: detection, evidence, and consequences.
Conservation Ecology, 5, 1.
Cane, J., Schiffhauer, D. & Kervin, L.J. (1996) Pollination, foraging, and nesting ecology
of the leaf-cutting bee Megachile (Delomegachile) addenda (Hymenoptera:
Megachilidae) on cranberry beds. Annals of the Entomological Society of America,
361–367.
Carreck, N. & Williams, I. (2002) Food for insect pollinators on farmland: insect visits to
flowers of annual seed mixtures. Journal of Insect Conservation, 6, 13–23.
Carroll, A., Pallardy, S. & Galen, C. (2001) Drought stress, plant water status, and floral
trait expression in fireweed, Epilobium angustifolium (Onagraceae). American
Journal of Botany, 88, 438–446.
Carvalheiro, L.G., Seymour, C.L., Nicolson, S.W. & Veldtman, R. (2012) Creating
patches of native flowers facilitates crop pollination in large agricultural fields:
mango as a case study. Journal of Applied Ecology, 49, 1373–1383.
Carvell, C., Meek, W.R., Pywell, R.F., Goulson, D. & Nowakowski, M. (2007)
Comparing the efficacy of agri-environment schemes to enhance bumble bee
abundance and diversity on arable field margins. Journal of Applied Ecology, 44,
29–40.
Carvell, C., Roy, D.B., Smart, S.M., Pywell, R.F., Preston, C.D. & Goulson, D. (2006a)
Declines in forage availability for bumblebees at a national scale. Biological
Conservation, 132, 481–489.
Carvell, C., Westrich, P., Meek, W.R., Pywell, R.F. & Nowakowski, M. (2006b)
Assessing the value of annual and perennial forage mixtures for bumblebees by
direct observation and pollen analysis. Apidologie, 37, 326–340.
Chao, A. (1987) Estimating the population size for capture-recapture data with unequal
catchability. Biometrics, 43, 783–791.
Chao, A. & Shen, T.J. (2003) Nonparametric estimation of Shannon’s index of diversity
when there are unseen species in sample. Environmental and Ecological Statistics,
10, 429–443.

150

Chapin, F.S., Zavaleta, E.S., Eviner, V.T., Naylor, R.L., Vitousek, P.M., Reynolds, H.L.,
et al. (2000) Consequences of changing biodiversity. Nature, 405, 234–42.
Chauzat, M.-P., Faucon, J.-P., Martel, A.-C., Lachaize, J., Cougoule, N. & Aubert, M.
(2006) A survey of pesticide residues in pollen loads collected by honey bees in
France. Journal of Economic Entomology, 99, 253–262.
Chavez, D.J. & Lyrene, P.M. (2009) Effects of self-pollination and cross-pollination of
Vaccinium darrowii (Ericaceae) and other low-chill blueberries. HortScience, 44,
1538–1541.
Clarke, K.R., Somerfield, P.J. & Chapman, M.G. (2006) On resemblance measures for
ecological studies, including taxonomic dissimilarities and a zero-adjusted Bray–
Curtis coefficient for denuded assemblages. Journal of Experimental Marine
Biology and Ecology, 330, 55–80.
Colla, S.R., Otterstatter, M.C., Gegear, R.J. & Thomson, J.D. (2006) Plight of the bumble
bee: Pathogen spillover from commercial to wild populations. Biological
Conservation, 129, 461–467.
Colla, S.R. & Packer, L. (2008) Evidence for decline in eastern North American
bumblebees (Hymenoptera: Apidae), with special focus on Bombus affinis
Cresson. Biodiversity and Conservation, 17, 1379–1391.
Colla, S.R. & Ratti, C.M. (2010) Evidence for the decline of the western bumble bee
(Bombus occidentalis Greene) in British Columbia. The Pan-Pacific Entomologist,
86, 32–34.
Cox, P. & Elmqvist, T. (2000) Pollinator extinction in the Pacific Islands. Conservation
Biology, 14, 1237–1239.
Crane, E. (1990) Bees and beekeeping: science, practice and world resources. Cornell
University Press. 604 p.
Cresswell, J.E., Osborne, J.L. & Goulson, D. (2000) An economic model of the limits to
foraging range in central place foragers with numerical solutions for bumblebees.
Ecological Entomology, 25, 249–255.
Cresswell, J.E., Page, C.J., Uygun, M.B., Holmbergh, M., Li, Y., Wheeler, J.G., et al.
(2012) Differential sensitivity of honey bees and bumble bees to a dietary
insecticide (imidacloprid). Zoology, 115, 365–371.
Cuthbertson, P.S. & Jepson, P.C. (1988) Reducing pesticide drift into the hedgerow by
the inclusion of an unsprayed field margin. Proceedings of the Brighton Crop
Protection Conference-Pests and Diseases, 2, 747-751.

151

Davis, B.N. & Williams, C.T. (1990) Buffer zone widths for honeybees from ground and
aerial spraying of insecticides. Environmental Pollution, 63, 247–259.
Davis, B.N.K., Lakhani, K.H. & Yates, T.J. (1991) The hazards of insecticides to
butterflies of field margins. Agriculture, Ecosystems & Environment, 36, 151–161.
Dedej, S. & Delaplane, K. (2003) Honey bee (Hymenoptera:Apidae) pollination of
rabbiteye blueberry Vaccinium ashei var. “Climax” is pollinator densitydependent. Horticultural Entomology, 96, 1215–1220.
Delaplane, K.S., Mayer, D.R. & Mayer, D.F. (2000) Crop Pollination by Bees. CABI.
352 p.
Desjardins, E.-C. & de Oliveira, D. (2006) Commercial bumble bee Bombus impatiens
(Hymenoptera: Apidae) as a pollinator of lowbush blueberry (Ericale: Ericaceae)
fields. Journal of Economic Entomology, 99, 443–449.
Desneux, N., Decourtye, A. & Delpuech, J.-M. (2007) The sublethal effects of pesticides
on beneficial arthropods. Annual Review of Entomology, 52, 81–106.
Deyrup, M., Edirisinghe, J. & Norden, B. (2002) The diversity and floral hosts of bees at
the Archbold Biological Station, Florida (Hymenoptera: Apoidea). Insecta Mundi,
16, 87–120.
DiPasquale, G., Salignon, M., Conte, Y. Le, Belzunces, L.P., Decourtye, A.,
Kretzschmar, A., et al. (2013) Influence of pollen nutrition on honey bee health:
Do pollen quality and diversity matter? PLoS ONE, 8.
Dormann, C.F., Schweiger, O., Augenstein, I., Bailey, D., Billeter, R., De Blust, G., et al.
(2007) Effects of landscape structure and land-use intensity on similarity of plant
and animal communities. Global Ecology and Biogeography, 16, 774–787.
Dorr, J. & Martin, E.C. (1966) Pollination studies on the highbush blueberry, Vaccinium
corymbosum L. Michigan Agricultural Experiment Station Quarterly Bulletin,
437–448.
Dover, J. & Sparks, T. (2000) A review of the ecology of butterflies in British
hedgerows. Journal of Environmental Management, 60, 51–63.
Droege, S., LeBuhn, G., Griswold, T.L., Minckley, R.L., Roulston, T., Cane, J.H., et al.
(2003) Standardized method for monitoring bee populations - the bee inventory
(BI) plot. White paper. Available: http://online.sfsu.edu/~beeplot.
Droege, S., Tepedino, V.J., LeBuhn, G., Link, W., Minckley, R.L., Chen, Q., et al. (2010)
Spatial patterns of bee captures in North American bowl trapping surveys. Insect
Conservation and Diversity, 3, 15–23.

152

Edwards, M. (1996) Optimizing habitats for bees in the United Kingdom - a review of
recent conservation action. In The Conservation of Bees, Linnean Society
Symposium Series No. 18 (ed. by Matheson, A., Buchmann, S.L., O’Toole, C.,
Westrich, P. & Williams, I.H.). Academic Press, London, pp. 35–45.
eFly: Spotted Wing Drosophila Working Group. (2012) Spotted wing drosophila impacts
in the eastern United States. White paper. Retrieved from:
http://www.sripmc.org/WorkingGroups/eFly/Impacts%20of%20SWD%20in%20th
e%20Eastern%20US%202012.pdf.
Elmqvist, T., Folke, C., Nyström, M., Peterson, G., Bengtsson, J., Walker, B., et al.
(2003) Response diversity, ecosystem change, and resilience. Frontiers in Ecology
and the Environment, 1, 488–494.
Elton, C.S. (1958) The ecology of invasions by animals and plants. Methuen, London,
UK. 196 p.
Fang, X., Turner, N.C., Yan, G., Li, F. & Siddique, K.H.M. (2010) Flower numbers, pod
production, pollen viability, and pistil function are reduced and flower and pod
abortion increased in chickpea (Cicer arietinum L.) under terminal drought.
Journal of Experimental Botany, 61, 335–345.
Fontaine, C., Dajoz, I., Meriguet, J. & Loreau, M. (2006) Functional diversity of plantpollinator interaction webs enhances the persistence of plant communities. PLoS
Biology, 4, 129–135.
Forrest, J., Thorp, R.W., Kremen, C. & Williams, N.M. (2015) Contrasting patterns in
species and functional-trait diversity of bees in an agricultural landscape. Journal
of Applied Ecology, early view. DOI: 10.1111/1365–2664.12433.
Francis, J.A. & Vavrus, S.J. (2012) Evidence linking Arctic amplification to extreme
weather in mid-latitudes. Geophysical Research Letters, 39, L06801, DOI:
10.1029/2012GL051000.
Free, J.B. (1993) Insect Pollination of Crops. Harcourt Brace Jovanovich, London. 544 p.
Fürst, M.A., McMahon, D.P., Osborne, J.L., Paxton, R.J. & Brown, M.J.F. (2014)
Disease associations between honeybees and bumblebees as a threat to wild
pollinators. Nature, 506, 364–366.
Galecki, A., & Burzykowski, T. (2013) Linear Mixed-Effect Models Using R: A Step-byStep Approach. Springer Texts in Statistics. Springer, New York. 542 p.
Garibaldi, L.A., Steffan-Dewenter, I., Winfree, R., Aizen, M.A., Bommarco, R.,
Cunningham, S.A., et al. (2013) Wild pollinators enhance fruit set of crops
regardless of honey bee abundance. Science, 339, 1608–1611.

153

Gathmann, A. & Tscharntke, T. (2002) Foraging ranges of solitary bees. Journal of
Animal Ecology, 71, 757–764.
Gezon, Z.J., Wyman, E.S., Ascher, J.S., Inouye, D.W. & Irwin, R.E. (2015) The effect of
repeated, lethal sampling on wild bee abundance and diversity. Methods in
Ecology and Evolution, early view. DOI: 10.1111/2041-210X.12375.
Gibbs, J. (2011) Revision of the metallic Lasioglossum (Dialictus) of eastern North
America (Hymenoptera: Halictidae: Halictini). Zootaxa, 3073, 1–216.
Giles, V. & Ascher, J.S. (2006) A survey of the bees of the Black Rock Forest Preserve,
New York (Hymenoptera: Apoidea). Journal of Hymenoptera Research, 15, 208–
231.
Gill, R.J., Ramos-Rodriguez, O. & Raine, N.E. (2012) Combined pesticide exposure
severely affects individual- and colony-level traits in bees. Nature, 491, 105–8.
Gonthier, D.J., Ennis, K.K., Farinas, S., Hsieh, H.-Y., Iverson, A.L., Batáry, P., et al.
(2014) Biodiversity conservation in agriculture requires a multi-scale approach.
Proceedings of the Royal Society of London B-Biological Sciences, 281, 20141358.
Gotelli, N.J. & Colwell, R.K. (2001) Quantifying biodiversity: Procedures and pitfalls in
the measurement and comparison of species richness. Ecology Letters, 4, 379-391.
Gotelli, N.J. & Ellison, A.M. (2004) A primer of ecological statistics. Sinauer Associates.
492 p.
Goulson, D. (2003) Effects of introduced bees on native ecosystems. Annual Review of
Ecology, Evolution, and Systematics, 34, 1–26.
Goulson, D., Lye, G.C. & Darvill, B. (2008) Decline and conservation of bumble bees.
Annual Review of Entomology, 53, 191–208.
Goulson, D., Nicholls, E., Botías, C. & Rotheray, E.L. (2015) Bee declines driven by
combined stress from parasites, pesticides, and lack of flowers. Science, early
view. DOI: 10.1126/science.1255957.
Greenleaf, S.S., Williams, N.M., Winfree, R. & Kremen, C. (2007) Bee foraging ranges
and their relationship to body size. Oecologia, 153, 589–596.
Grixti, J.C., Wong, L.T., Cameron, S.A. & Favret, C. (2009) Decline of bumble bees
(Bombus) in the North American Midwest. Biological Conservation, 142, 75–84.
Grundel, R., Jean, R.P., Frohnapple, K.J., Glowacki, G. a., Scott, P.E. & Pavlovic, N.B.
(2010) Floral and nesting resources, habitat structure, and fire influence bee

154

distribution across an open-forest gradient. Ecological Applications, 20, 1678–
1692.
Hauser, M. (2011) A historic account of the invasion of Drosophila suzukiii (Matsumara)
(Diptera: Drosophilidae) in the continental United States, with remarks on their
identification. Pest Management Science, 67, 1352–1357.
Haydak, M.H. (1970) Honey bee nutrition. Annual Review of Entomology, 15, 143–156.
Hegland, S.J., Nielsen, A., Lázaro, A., Bjerknes, A.-L. & Totland, Ø. (2009) How does
climate warming affect plant-pollinator interactions? Ecology Letters, 12, 184–95.
Heinrich, B. (2004) Bumblebee Economics. Harvard University Press, Cambridge, MA.
Helson, B.V., Barber, K.N. & Kingsbury, P.D. (1994) Laboratory toxicology of six
forestry insecticides to four species of bee (Hymenoptera: Apoidea). Archives of
Environmental Contamination and Toxicology, 27, 107–114.
Hendrickx, F., Maelfait, J.P., Wingerden, W. Van, Schweiger, O., Speelmans, M.,
Aviron, S., et al. (2007) How landscape structure, land-use intensity and habitat
diversity affect components of total arthropod diversity in agricultural landscapes.
Journal of Applied Ecology, 44, 340–351.
Hernandez, J.L., Frankie, G.W. & Thorp, R.W. (2009) Ecology of urban bees: A review
of current knowledge and directions for future study. Cities and the Environment,
2, 1–15.
Hines, H.M. & Hendrix, S.D. (2005) Bumble bee (Hymenoptera!: Apidae) diversity and
abundance in tallgrass prairie patches: Effects of local and landscape floral
resources. Environmental Entomology, 34, 1477–1484.
Hoehn, P., Tscharntke, T., Tylianakis, J.M. & Steffan-Dewenter, I. (2008) Functional
group diversity of bee pollinators increases crop yield. Proceedings of the Royal
Society B-Biological Sciences, 275, 2283–2291.
Hokanson, K. & Hancock, J. (2000) Early-acting inbreeding depression in three species
of Vaccinium (Ericaceae). Sexual Plant Reproduction, 13, 145–150.
Holland, J.D., Bert, D.G. & Fahrig, L. (2004) Determining the spatial scale of species’
response to habitat. BioScience, 54, 227–233.
Hooper, D., Chapin, F., Ewel, J., Hector, A., Inchausti, P., Lavorel, S., et al. (2005)
Effects of biodiversity on ecosystem functioning: A consensus of current
knowledge. Ecological Monographs, 75, 3–35.

155

Hoover, S.E.R., Higo, H.A. & Winston, M.L. (2006) Worker honey bee ovary
development: seasonal variation and the influence of larval and adult nutrition.
Journal of Comparative Physiology. B, Biochemical, Systemic, and Environmental
Physiology, 176, 55–63.
Huang, Z. (2012) Pollen nutrition affects honey bee stress resistance. Terrestrial
Arthropod Reviews, 5, 175–189.
Isaacs, R. & Kirk, A.K. (2010) Pollination services provided to small and large highbush
blueberry fields by wild and managed bees. Journal of Applied Ecology, 47, 841–
849.
Isaacs, R., Wise, J.C. & Garcia-Salazar, C. (2009) Developing integrated fruitworm
control strategies for blueberry in preparation for pesticide restrictions. Acta
Horticulturae, 810, 259–264.
Iwasa, T., Motoyama, N., Ambrose, J.T. & Roe, R.M.M. (2004) Mechanism for the
differential toxicity of neonicotinoid insecticides in the honey bee, Apis mellifera.
Crop Protection, 23, 371–378.
Jackson, D.A. (1993) Stopping rules in principal components analysis - a comparison of
heuristic and statistical approaches. Ecology, 74, 2204–2214.
Javorek, S.K., MacKenzie, K.E. & VanderKloet, S.P. (2002) Comparative pollination
effectiveness among bees (Hymenoptera: Apoidea) on lowbush blueberry
(Ericaceae: Vaccinium angustifolium). Annals of the Entomological Society of
America, 95, 345–351.
Jax, K. (2010) Ecosystem functioning. Cambridge University Press, Cambridge, UK. 286
p.
Jha, S. & Kremen, C. (2013a) Resource diversity and landscape-level homogeneity drive
native bee foraging. Proceedings of the National Academy of Sciences of the
United States of America, 110, 555–8.
Jha, S. & Kremen, C. (2013b) Urban land use limits regional bumble bee gene flow.
Molecular Ecology, 22, 2483–2495.
Johansen, C.A. (1977) Pesticides and pollinators. Annual Review of Entomology, 22,
177–192.
Johnson, R.M., Ellis, M.D., Mullin, C.A. & Frazier, M. (2010) Pesticides and honey bee
toxicity – USA. Apidologie, 41, 312–331.

156

Jönsson, A.M., Ekroos, J., Dänhardt, J., Andersson, G.K.S., Olsson, O. & Smith, H.G.
(2015) Sown flower strips in southern Sweden increase abundances of wild bees
and hoverflies in the wider landscape. Biological Conservation, 184, 51–58.
Karl, T.R., Gleason, B.E., Menne, M.J., McMahon, J.R., Heim, R.R., Brewer, M.J., et al.
(2012) U.S. temperature and drought: Recent anomalies and trends. Eos, 93, 473–
474.
Kassambara, A. (2015) factoextra: Visualization of the outputs of a multivariate analysis.
R package 1.0.1. Available: http://www.sthda.com.
Katz, R.W. & Brown, B.G. (1992) Extreme events in a changing climate: Variability is
more important than averages. Climatic Change, 21, 289–302.
Kearns, C.A. & Inouye, D.W. (1997) Pollinators, flowering plants, and conservation
biology. BioScience, 47, 297–307.
Kearns, C.A., Inouye, D.W. & Waser, N.M. (1998) Endangered mutualisms: the
conservation of plant-pollinator interactions. Annual Review of Ecology and
Systematics, 29, 83–112.
Keddy, P.A. (1992) Assembly and response rules: two goals for predictive community
ecology. Journal of Vegetation Science, 3 (2), 157-164.
Kells, A.R., Holland, J.M. & Goulson, D. (2001) The value of uncropped field margins
for foraging bumblebees. Journal of Insect Conservation, 5, 283–291.
Kennedy, C.M., Lonsdorf, E., Neel, M.C., Williams, N.M., Ricketts, T.H., Winfree, R., et
al. (2013) A global quantitative synthesis of local and landscape effects on wild
bee pollinators in agroecosystems. Ecology Letters, 16, 584–599.
Kevan, P.G. (1975) Forest application of the insecticide fenitrothion and its effect on wild
bee pollinators (Hymenoptera: Apoidea) of lowbush blueberries (Vaccinium spp.)
in Southern New Brunswick, Canada. Biological Conservation, 7, 301–309.
Kevan, P.G. (1977) Blueberry crops in Nova Scotia and New Brunswick - pesticides and
crop reductions. Canadian Journal of Agricultural Economics, 25, 61–64.
Kevan, P.G., Greco, C.F. & Belaoussoff, S. (1997) Log-normality of biodiversity and
abundance in diagnosis and measuring of ecosystemic health: pesticide stress on
pollinators on blueberry heaths. Journal of Applied Ecology, 34, 1122–1136.
Kim, J., Williams, N.M. & Kremen, C. (2006) Effects of cultivation and proximity to
natural habitat on ground-nesting native bees in California sunflower fields.
Journal of the Kansas Entomological Society, 79, 309–320.

157

Kleijn, D. & Raemakers, I. (2008) A retrospective analysis of pollen host plant use by
stable and declining bumble bee species. Ecology, 89, 1811–1823.
Kleijn, D. & van Langevelde, F. (2006) Interacting effects of landscape context and
habitat quality on flower visiting insects in agricultural landscapes. Basic and
Applied Ecology, 7, 201–214.
Klein, A.-M., Steffan-Dewenter, I. & Tscharntke, T. (2003) Fruit set of highland coffee
increases with the diversity of pollinating bees. Proceedings of the Royal Society of
London B-Biological Sciences, 270, 955–961.
Klein, A.-M., Vaissiere, B.E., Cane, J.H., Steffan-Dewenter, I., Cunningham, S.A.,
Kremen, C., et al. (2007) Importance of pollinators in changing landscapes for
world crops. Proceedings of the Royal Society of London B-Biological Sciences,
274, 303–313.
Kraft, N.J.B., Adler, P.B., Godoy, O., James, E., Fuller, S. & Levine, J.M. (2014)
Community assembly, coexistence, and the environmental filtering metaphor.
Functional Ecology, early view. DOI: 10.1111/1365–2435.12345.
Krebs, J.R., Wilson, J.D., Bradbury, R.B. & Siriwardena, G.M. (1999) The second Silent
Spring? Nature, 400, 611–612.
Krebs, S.L. & Hancock, J.F. (1990) Early-acting inbreeding depression and reproductive
success in the highbush blueberry, Vaccinium corymbosum L. Theoretical and
Applied Genetics, 79, 825–32.
Kremen, C. (2005) Managing ecosystem services: what do we need to know about their
ecology? Ecology Letters, 8, 468–479.
Kremen, C., Williams, N.M., Aizen, M.A., Gemmill-Herren, B., LeBuhn, G., Minckley,
R., et al. (2007) Pollination and other ecosystem services produced by mobile
organisms: a conceptual framework for the effects of land-use change. Ecology
Letters, 10, 299–314.
Kremen, C., Williams, N.M. & Thorp, R.W. (2002) Crop pollination from native bees at
risk from agricultural intensification. Proceedings of the National Academy of
Sciences of the United States of America, 99, 16812–6.
Kreyer, D., Oed, A., Walther-Hellwig, K. & Frankl, R. (2004) Are forests potential
landscape barriers for foraging bumblebees? Landscape scale experiments with
Bombus terrestris and Bombus pascuorum (Hymenoptera, Apidae). Biological
Conservation, 116, 111–118.
Krombein, K. V. (1967) Trap-nesting wasps and bees: Life histories, nests and
associates. Smithsonian Institution Press, Washington, D.C. 570 pp.

158

Krupke, C.H., Hunt, G.J., Eitzer, B.D., Andino, G. & Given, K. (2012) Multiple routes of
pesticide exposure for honey bees living near agricultural fields. PLoS ONE, 7, 1–
8.
LaBerge, W.E. (1980) A revision of the bees of the genus Andrena of the western
hemisphere. Part X. Subgenus Andrena. Transactions of the American
Entomological Society, 106, 395–525.
Laird, N.M. & Ware, J.H. (1982) Random-effects models for longitudinal data.
Biometrics, 38, 963–974.
Laliberte, E., Legendre, P., Ecology, S. & January, N. (2010) A distance-based
framework for measuring functional diversity from multiple traits. Ecology, 91,
299–305.
Lang, G.A. & Danka, R.G. (1991) Honey-bee-mediated cross- versus self-pollination of
“Sharpblue” blueberry increases fruit size and hastens ripening. Journal of the
American Society for Horticultural Science, 116, 770–773.
Larsen, T.H., Williams, N.M. & Kremen, C. (2005) Extinction order and altered
community structure rapidly disrupt ecosystem functioning. Ecology Letters, 8,
538–47.
Lavelle, P. (1996) Diversity of soil fauna and ecosystem function. Biology International,
33, 3–16.
Lavorel, S., & Garnier, E. (2002) Predicting changes in community composition and
ecosystem functioning from plant traits: revisiting the Holy Grail. Functional
Ecology, 16, 545-556.
Lebuhn, G., Droege, S., Connor, E.F., Gemmill-Herren, B., Potts, S.G., Minckley, R.L.,
et al. (2013) Detecting insect pollinator declines on regional and global scales.
Conservation Biology, 27, 113–120.
Lebuhn, G., Droege, S., Connor, E.F., Gemmill-Herren, B., Potts, S.G., Minckley, R.L.,
et al. (2015) Evidence-based conservation: reply to Tepedino et al. Conservation
Biology, 29, 283–285.
Le Féon, V., Schermann-Legionnet, A., Delettre, Y., Aviron, S., Billeter, R., Bugter, R.,
et al. (2010) Intensification of agriculture, landscape composition and wild bee
communities: A large scale study in four European countries. Agriculture,
Ecosystems and Environment, 137, 143–150.
Lee, J.C., Bruck, D.J., Curry, H., Edwards, D., Haviland, D.R., Steenwyk, R.A. van, et al.
(2011) The susceptibility of small fruits and cherries to the spotted-wing
drosophila (Drosophila suzukii). Pest Management Science, 67, 1358–1367.

159

Lever, J.J., van Nes, E.H., Scheffer, M. & Bascompte, J. (2014) The sudden collapse of
pollinator communities. Ecology Letters, 17, 350–9.
Longley, M., Čilgi, T., Jepson, P.C. & Sotherton, N.W. (1997) Measurements of
pesticide spray drift deposition into field boundaries and hedgerows: 1. Summer
applications. Environmental Toxicology and Chemistry, 16, 165–172.
Longley, M. & Sotherton, N.W. (1997) Factors determining the effects of pesticides upon
butterflies inhabiting arable farmland. Agriculture, Ecosystems and Environment,
61, 1–12.
Lonsdorf, E., Kremen, C., Ricketts, T., Winfree, R., Williams, N. & Greenleaf, S. (2009)
Modelling pollination services across agricultural landscapes. Annals of Botany,
103, 1589–600.
Lonsdorf, E., Ricketts, T., Kremen, C., Winfree, R., Greenleaf, S. & Williams, N. (2011)
Crop pollination services. In Natural Capital: Theory and practice of mapping
ecosystem services. (ed. by Karvera, P., Tallis, H., Ricketts, T.H., Daily, G.C. &
Polasky, S.). Oxford University Press, Oxford, UK., pp. 168–187.
Loreau, M., Mouquet, N. & Gonzales, A. (2003) Biodiversity as spatial insurance in
heterogeneous landscapes. Proceedings of The National Academy of Sciences of
the United States of America, 100, 12765–12770.
Loreau, M., Naeem, S. & Inchausti, P. (2002) Biodiversity and ecosystem functioning:
Synthesis and perspectives. Oxford University Press, Oxford, UK.
Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J.P., Hector, a, et al. (2001)
Biodiversity and ecosystem functioning: current knowledge and future challenges.
Science, 294, 804–808.
Luck, G.W., Harrington, R., Harrison, P. a, Kremen, C., Berry, P.M., Bugter, R., et al.
(2009) Quantifying the contribution of organisms to the provison of ecosystem
services. BioScience, 59, 223–235.
MacArthur, R. (1955) Fluctuations of animal populations and a measure of community
stability. Ecology, 36, 533–536.
MacKenzie, K. (2009) Pollination practices and the use of bees in Vaccinium crops. Acta
Horticulturae, 810, 527–538.
MacKenzie, K.E. & Eickwort, G.C. (1996) Diversity and abundance of bees
(Hymenoptera: Apoidea) foraging on highbush blueberry (Vaccinium corymbosum
L.) in central New York. Journal of the Kansas Entomological Society, 69, 185–
194.

160

Mandelik, Y., Winfree, R., Neeson, T. & Kremen, C. (2012) Complementary habitat use
by wild bees in agro-natural landscapes. Ecological Applications, 22, 1535–1546.
Marshall, E.J.P., West, T.M. & Kleijn, D. (2006) Impacts of an agri-environment field
margin prescription on the flora and fauna of arable farmland in different
landscapes. Agriculture Ecosystems & Environment, 113, 36–44.
Matson, P.A., Parton, W.J., Power, A.G. & Swift, M.J. (1997) Agricultural intensification
and ecosystem properties. Science, 277, 504–509.
May, R.M. (1973) Stability and complexity in model ecosystems. Monographs in
Population Biology, 6, 1–235.
McGregor, S.E. (1976) Insect Pollination of Cultivated Crop Plants. Agriculture
Handbook No. 496. Agricultural Research Service, USDA, Washington, D.C.
Meader, E.M. & Darrow, G.M. (1947) Highbush blueberry pollination experiments.
Proceedings of the American Society of Horticultural Science, 49, 196–204.
Memmott, J., Waser, N.M. & Price, M. V. (2004) Tolerance of pollination networks to
species extinctions. Proceedings of the Royal Society of London B-Biological
Sciences, 271, 2605–2611.
Merrill, T.A. (1936) Pollination of highbush blueberry. Michigan Agricultural
Experiment Station Techical Bulletin 151.
Michener, C.D. (1979) Biogeography of the bees. Annals of the Missouri Botanical
Garden, 66, 278–347.
Michener, C.D. (2000) The Bees of the World. The Johns Hopkins University Press,
Baltimore, MD.
Michener, C.D., Lange, R.B., Bigarella, J.J. & Salamuni, R. (1958) Factors influencing
the distribution of bees’ nests in earth banks. Ecology, 39, 207–217.
Michener, C.D., McGinley, R.J. & Danforth, B.N. (1994) The Bee Genera of North and
Central America (Hymenoptera: Apoidea). Smithsonian Institution Press,
Washington, D.C.
Minckley, R.L., Roulston, T.H. & Williams, N.M. (2013) Resource assurance predicts
specialist and generalist bee activity in drought. Proceedings of the Royal Society
B-Biological Sciences, 280, 20122703.
Mitchell, T.B. (1960) Bees of the Eastern United States. Volume 1. Technical Bulletin
No. 141. North Carolina Agricultural Experiment Station, Raleigh, NC.

161

Mitchell, T.B. (1962) Bees of the Eastern United States. Volume 2. Technical Bulletin
No. 152. North Carolina Agricultural Experiment Station, Raleigh, NC.
Moisan-Deserres, J., Girard, M., Chagnon, M. & Fournier, V. (2014) Pollen loads and
specificity of native pollinators of lowbush blueberry. Journal of Economic
Entomology, 107, 1156–1162.
Morandin, L.A., Winston, M.L., Franklin, M.T. & Abbott, V.A. (2005) Lethal and sublethal effects of spinosad on bumble bees (Bombus impatiens Cresson). Pest
Management Science, 61, 619–626.
Moritz, R.F.A., Härtel, S. & Neumann, P. (2005) Global invasions of the western
honeybee (Apis mellifera) and the consequences for biodiversity. Ecoscience, 12,
289–301.
Mullin, C.A., Frazier, M., Frazier, J.L., Ashcraft, S., Simonds, R., VanEngelsdorp, D., et
al. (2010) High levels of miticides and agrochemicals in North American apiaries:
implications for honey bee health. PLoS ONE, 5, 1–19.
Murray, T., Kuhlmann, M. & Potts, S.G. (2009) Conservation ecology of bees:
populations, species and communities. Apidologie, 40, 211–236.
Naeem, S. (1998) Species redundancy and ecosystem reliability. Conservation Biology,
12, 39–45.
Naeem, S., Bunker, D.E., Hector, A., Loreau, M., Perrings, C., Schmid, B., et al. (2009)
Biodiversity, Ecosystem Functioning , and Human Wellbeing: An Ecological and
Economic Perspective. Oxford University Press, Oxford, UK.
National Agricultural Statistics Service [NASS]. (2001-11). Agricultural Chemical
Usage: Fruit Crops (2000-10). Agricultural Statistics Board, United States
Department of Agriculture, Washington, D.C.
National Agricultural Statistics Service [NASS]. (2014) Noncitrus Fruits and Nuts: 2013
Summary. United States Department of Agriculture, Washington, D.C.
Oksanen, J. (2006) Multivariate Analysis of Ecological Communities in R: vegan tutorial,
39. http://cc.oulu.fi/~jarioksa/opetus/metodi/vegantutor.pdf (Accessed February
2014).
Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., Minchin, P.R., O’Hara, R.B., et al.
(2013) vegan: Community Ecology Package. R package version 2.0-10. Available:
http://CRAN.R-project.org/package=vegan.
Ollerton, J., Winfree, R. & Tarrant, S. (2011) How many flowering plants are pollinated
by animals? Oikos, 120, 321–326.

162

Osborne, J.L., Martin, A.P., Carreck, N.L., Swain, J.L., Knight, M.E., Goulson, D., et al.
(2008) Bumblebee flight distances in relation to the forage landscape. Journal of
Animal Ecology, 77, 406–415.
Osgood, E.A. (1972) Soil characteristics of nesting sites of solitary bees associated with
the lowbush blueberry in Maine. University of Maine at Orono Technical Bulletin
59.
Otterstatter, M.C. & Thomson, J.D. (2008) Does pathogen spillover from commercially
reared bumble bees threaten wild pollinators? PLoS ONE, 3, 1–9.
Otti, O. & Schmid-Hempel, P. (2007) Nosema bombi: A pollinator parasite with
detrimental fitness effects. Journal of Invertebrate Pathology, 96, 118–24.
Otti, O. & Schmid-Hempel, P. (2008) A field experiment on the effect of Nosema bombi
in colonies of the bumblebee Bombus terrestris. Ecological Entomology, 33, 577–
582.
Otto, S., Lazzaro, L., Finizio, A. & Zanin, G. (2009) Estimating ecotoxicological effects
of pesticide drift on nontarget arthropods in field hedgerows. Environmental
Toxicology and Chemistry, 28, 853–863.
Pawar, S. (2014) Why are plant-pollinator networks nested? Science, 345, 383–383.
Petanidou, T., Kallimanis, A.S., Tzanopoulos, J., Sgardelis, S.P. & Pantis, J.D. (2008)
Long-term observation of a pollination network: fluctuation in species and
interactions, relative invariance of network structure and implications for estimates
of specialization. Ecology Letters, 11, 564–75.
Pettis, J.S., VanEngelsdorp, D., Johnson, J. & Dively, G. (2012) Pesticide exposure in
honey bees results in increased levels of the gut pathogen Nosema. Die
Naturwissenschaften, 99, 153–158.
Pilling, E.D. & Jepson, P.C. (1993) Synergism between EBI fungicides and a pyrethroid
insecticide in the honeybee (Apis mellifera). Pesticide Science, 39, 293–297.
Pimm, S.L. (1979) Complexity and stability: another look at MacArthur’s original
hypothesis. Oikos, 33, 351–357.
Pimm, S.L. (1984) The complexity and stability of ecosystems. Nature, 307, 321–326.
Pimm, S.L., Russell, G.J., Gittleman, J.L. & Brooks, T.M. (1995) The future of
biodiversity. Science, 269, 347–350.

163

Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. & Team, R.C. (2014) nlme: Linear and
Nonlinear Mixed Effects Models. R package version 3.1-118. Available:
http://CRAN.R-project.org/package=nlme.
Plowright, R.C. & Rodd, F.H. (1980) The effect of aerial insecticide spraying on
hymenopterous pollinators in New Brunswick. The Canadian Entomologist, 112,
259–269.
Potts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger, O. & Kunin, W.E.
(2010) Global pollinator declines: trends, impacts and drivers. Trends in Ecology
& Evolution, 25, 345–353.
Potts, S.G., Vulliamy, B., Dafni, A., Ne’eman, G. & Willmer, P. (2003) Linking bees and
flowers: how do floral communities structure pollinator communities? Ecology, 84,
2628–2642.
Potts, S.G., Vulliamy, B., Roberts, S., O’Toole, C., Dafni, A., Ne’eman, G., et al. (2005)
Role of nesting resources in organizing diverse bee communities in a
Mediterranean landscape. Ecological Entomology, 30, 78–85.
Potts, S.G. & Willmer, P. (1997) Abiotic and biotic factors influencing nest-site selection
by Halictus rubicundus, a ground-nesting halictine bee. Ecological Entomology,
22, 319–328.
Pywell, R.F., Warman, E.A., Hulmes, L., Hulmes, S., Nuttall, P., Sparks, T.H., et al.
(2007) Effectiveness of new agri-environment schemes in providing foraging
resources for bumblebees in intensively farmed landscapes. Biological
Conservation, 129, 192–206.
Ratti, C.M., Higo, H.A., Griswold, T.L. & Winston, M.L. (2012) Bumble bees influence
berry size in commercial Vaccinium spp. cultivation in British Columbia. The
Canadian Entomologist, 140, 348–363.
Ravoet, J., Smet, L. De, Meeus, I., Smagghe, G., Wenseleers, T. & de Graaf, D.C. (2014)
Widespread occurrence of honey bee pathogens in solitary bees. Journal of
Invertebrate Pathology, 122, 55–8.
Rehan, S.M., Rotella, A., Onuferko, T.M. & Richards, M.H. (2013) Colony disturbance
and solitary nest initiation by workers in the obligately eusocial sweat bee,
Halictus ligatus. Insectes Sociaux, 60, 389–392.
Rehan, S.M. & Sheffield, C.S. (2011) Morphological and molecular delineation of a new
species in the Ceratina dupla species-group (Hymenoptera: Apidae: Xylocopinae)
of eastern North America. Zootaxa, 50, 35–50.

164

Richards, M.H., Packer, L. & Seger, J. (1995) Unexpected patterns of parentage and
relatedness in a primitively eusocial bee. Nature, 373, 239–241.
Ricketts, T.H., Regetz, J., Steffan-Dewenter, I., Cunningham, S. a., Kremen, C.,
Bogdanski, A., et al. (2008) Landscape effects on crop pollination services: Are
there general patterns? Ecology Letters, 11, 499–515.
Rogers, S., Tarpy, D. & Burrack, H. (2013) Multiple criteria for evaluating pollinator
performance in highbush blueberry (Ericales!: Ericaceae) agroecosystems.
Environmental Entomology, 42, 1201–1209.
Rogers, S., Tarpy, D. & Burrack, H. (2014) Bee species diversity enhances productivity
and stability in a perennial crop. PLoS ONE, 9, 1–8.
Rohr, R.P., Saavedra, S. & Bascompte, J. (2014) On the structural stability of mutualistic
systems. Science, 345, 1253497–1253497.
Rosenfeld, J. (2002) Functional redundancy in ecology and conservation. Oikos, 98, 156–
162.
Roubik, D.W. (2001) Ups and downs in pollinator populations: when is there a decline?
Conservation Ecology, 5, 2.
Roulston, T. & Goodell, K. (2011) The role of resources and risks in regulating wild bee
populations. Annual Review of Entomology, 56, 293–312.
Rucker, R.R., Thurman, W.N. & Burgett, M. (2012) Honey bee pollination markets and
the internalization of reciprocal benefits. American Journal of Agricultural
Economics, 94, 956–977.
Sala, O.E., Chapin, F.S., Armesto, J.J., Berlow, E., Bloomfield, J., Dirzo, R., et al. (2000)
Global biodiversity scenarios for the year 2100. Science, 287, 1770–1774.
Sampson, B.J. & Cane, J.H. (2000) Pollination efficiencies of three bee (Hymenoptera:
Apoidea) species visiting rabbiteye blueberry. Journal of Economic Entomology,
93, 1726–1731.
Sardiñas, H.S. & Kremen, C. (2014) Evaluating nesting microhabitat for ground-nesting
bees using emergence traps. Basic and Applied Ecology, 15, 161–168.
Schaub, I.O. & Baver, L.D. (1942) Blueberries earlier and larger with cross-pollination.
Report of the North Carolina Agricultural Experiment Station, 65, 53.
Schweiger, O., Biesmeijer, J.C., Bommarco, R., Hickler, T., Hulme, P.E., Klotz, S., et al.
(2010) Multiple stressors on biotic interactions: how climate change and alien
species interact to affect pollination. Biological Reviews, 85, 777–95.

165

Scott-Dupree, C.D., Conroy, L. & Harris, C.R. (2009) Impact of currently used or
potentially useful insecticides for canola agroecosystems on Bombus impatiens
(Hymenoptera: Apidae), Megachile rotundata (Hymenoptera: Megachilidae), and
Osmia lignaria (Hymenoptera: Magachilidae). Journal of Economic Entomology,
102, 177–182.
Sgolastra, F., Kemp, W.P., Buckner, J.S., Pitts-Singer, T.L., Maini, S. & Bosch, J. (2011)
The long summer: Pre-wintering temperatures affect metabolic expenditure and
winter survival in a solitary bee. Journal of Insect Physiology, 57, 1651–1659.
Sheffield, C.S., Kevan, P.G., Smith, R.F., Rigby, S.M. & Rogers, R.E.L. (2003) Bee
species of Nova Scotia, Canada, with new records and notes on bionomics and
floral relations (Hymenoptera: Apoidea). Journal of the Kansas Entomological
Society, 76, 357–384.
Shuler, R.E., Roulston, T.H. & Farris, G.E. (2005) Farming practices influence wild
pollinator populations on squash and pumpkin. Journal of Economic Entomology,
98, 790–795.
Singh, R., Levitt, A.L., Rajotte, E.G., Holmes, E.C., Ostiguy, N., VanEngelsdorp, D., et
al. (2010) RNA viruses in hymenopteran pollinators: evidence of inter-taxa virus
transmission via pollen and potential impact on non-Apis hymenopteran species.
PLoS ONE, 5, 1–16.
Steffan-Dewenter, I. & Kuhn, A. (2003) Honeybee foraging in differentially structured
landscapes. Proceedings of the Royal Society of London Series B-Biological
Sciences, 270, 569–575.
Steffan-Dewenter, I., Munzenberg, U., Burger, C., Thies, C. & Tscharntke, T. (2002)
Scale-dependent effects of landscape context on three pollinator guilds. Ecology,
83, 1421–1432.
Steffan-Dewenter, I. & Tscharntke, T. (1999) Effects of habitat isolation on pollinator
communities and seed set. Oecologia, 121, 432–440.
Stout, J. & Morales, C. (2009) Ecological impacts of invasive alien species on bees.
Apidologie, 40, 388–409.
Stubbs, C.S. & Drummond, F.A. (1997) Pollination of wild lowbush blueberry,
Vaccinium angustifolium by the alfalfa leafcutting bee, Megachile rotundata . Acta
Horticulturae, 446, 189–196.
Stubbs, C.S. & Drummond, F.A. (2001) Bombus impatiens (Hymenoptera: Apidae): an
alternative to Apis mellifera (Hymenoptera: Apidae) for lowbush blueberry
pollination. Journal of Economic Entomology, 94, 609–616.

166

Stubbs, C.S., Jacobson, H.A., Osgood, E.A. & Drummond, F.A. (1992) Alternative
forage plants for native (wild) bees associated with lowbush blueberry, Vaccinium
spp., in Maine. Maine Agricultural Experiment Station, Technical Bulletin 148.
Svensson, B., Lagerlof, J. & Svensson, G. (2000) Habitat preferences of nest-seeking
bumble bees (Hymenoptera: Apidae) in an agricultural landscape. Agriculture,
Ecosystems & Environment, 77, 247–255.
Sydenham, M.A.K., Moe, S.R., Totland, Ø. & Eldegard, K. (2014) Does multi-level
environmental filtering determine the functional and phylogenetic composition of
wild bee species assemblages? Ecography, 38, 140-153.
Szabo, N.D., Colla, S.R., Wagner, D.L., Gall, L.F. & Kerr, J.T. (2012) Do pathogen
spillover, pesticide use, or habitat loss explain recent North American bumblebee
declines? Conservation Letters, 5, 232–239.
R Core Team (2013) R: A language and environment for statistical computing. R
Foundation for Statistical Computing, Vienna, Austria. Available: http://www.Rproject.org/.
Tepedino, V.J., Durham, S., Cameron, S.A. & Goodell, K. (2015) Documenting bee
decline or squandering scarce resources. Conservation Biology, 29, 280–282.
Tepedino, V. & Stanton, N. (1981) Diversity and competition in bee-plant communities
on short-grass prairie. Oikos, 36, 35–44.
Thompson, H. (2001) Assessing the exposure and toxicity of pesticides to bumblebees
(Bombus sp.). Apidologie, 32, 305–321.
Thompson, H. (2003) Behavioural effects of pesticides in bees: their potential for use in
risk assessment. Ecotoxicology, 12, 317–330.
Thompson, H. & Wilkins, S. (2003) Assessment of the synergy and repellency of
pyrethroid/fungicide mixtures. Bulletin of Insectology, 56, 131–134.
Thornton, P.K., Ericksen, P.J., Herrero, M. & Challinor, A.J. (2014) Climate variability
and vulnerability to climate change: a review. Global Change Biology, 11, 3313–
28.
Tscharntke, T., Gathmann, A. & Steffan-Dewenter, I. (1998) Bioindication using trapnesting bees and wasps and their natural enemies: community structure and
interactions. Journal of Applied Ecology, 35, 708–719.
Tscharntke, T., Klein, A.M., Kruess, A., Steffan-Dewenter, I. & Thies, C. (2005)
Landscape perspectives on agricultural intensification and biodiversity - ecosystem
service management. Ecology Letters, 8, 857–874.

167

Tuell, J.K. (2007) Ecology and conservation of wild bees associated with highbush
blueberry farms in Michigan. Dissertation. Michigan State University, East
Lansing, MI.
Tuell, J.K., Ascher, J.S. & Isaacs, R. (2009) Wild bees (Hymenoptera: Apoidea:
Anthophila) of the Michigan highbush blueberry agroecosystem. Annals of the
Entomological Society of America, 102, 275–287.
Tuell, J.K. & Isaacs, R. (2009) Elevated pan traps to monitor bees in flowering crop
canopies. Entomologia Experimentalis et Applicata, 131, 93–98.
Tuell, J.K. & Isaacs, R. (2010a) Community and species-specific responses of wild bees
to insect pest control programs applied to a pollinator-dependent crop. Journal of
Economic Entomology, 103, 668–675.
Tuell, J.K. & Isaacs, R. (2010b) Weather during bloom affects pollination and yield of
highbush blueberry. Journal of Economic Entomology, 103, 557–562.
Vanbergen, A.J. & Insect Pollinators Initiative. (2013) Threats to an ecosystem service:
Pressures on pollinators. Frontiers in Ecology and the Environment, 11, 251–259.
VanderKloet, S.P. (1991) The consequences of mixed pollination on seed set in
Vaccinium corymbosum. Canadian Journal of Botany-Revue Canadienne De
Botanique, 69, 2448–2454.
Van Timmeren, S., & Isaacs, R. (2013) Control of spotted wing drosophila, Drosophila
suzukii, by specific insecticides and by conventional and organic crop protection
programs. Crop Protection, 54, 126-133.
Vázquez, D.P. & Aizen, M.A. (2004) Asymmetric specialization: A pervasive feature of
plant-pollinator interactions. Ecology, 85, 1251–1257.
Walsh, D.B., Bolda, M.P., Goodhue, R.E., Dreves, A.J., Lee, J., Bruck, D.J., et al. (2011)
Drosophila suzukii (Diptera: Drosophilidae): Invasive pest of ripening soft fruit
expanding its geographic range and damage potential. Journal of Integrated Pest
Management, 2, 1–7.
Walther-Hellwig, K. & Frankl, R. (2000) Foraging habitats and foraging distances of
bumblebees, Bombus spp. (Hym., Apidae), in an agricultural landscape. Journal of
Applied Entomology, 124, 299–306.
Watson, J.C., Wolf, A.T. & Ascher, J.S. (2011) Forested landscapes promote richness
and abundance of native bees (Hymenoptera: Apoidea: Anthophila) in Wisconsin
apple orchards. Environmental Entomology, 40, 621–32.

168

Westphal, C., Bommarco, R., Carré, G., Lamborn, E., Morison, N., Petanidou, T., et al.
(2008) Measuring bee diversity in different European habitats and biogeographical
regions. Ecological Monographs, 78, 653–671.
Westphal, C., Steffan-Dewenter, I. & Tscharntke, T. (2006) Foraging trip duration of
bumblebees in relation to landscape-wide resource availability. Ecological
Entomology, 31, 389–394.
Westrich, P. (1996) Habitat requirements of central European bees and the problems of
partial habitats. In The Conservation of Bees, Linnean Society Symposium Series
No. 18 (ed. by Matheson, A., Buchmann, S.L., O’Toole, C., Westrich, P. &
Williams, I.H.). Academic Press, London, pp. 1–16.
Westwood, L. (2006) Habitat use by wild bees in a human-altered landscape.
Dissertation. Princeton University, Princeton, NJ.
Williams, I.H. (1994) The dependence of crop production within the European Union on
pollination by honey bees. Agricultural Zoology Reviews, 6, 229–257.
Williams, N.M., Crone, E.E., Roulston, T.H., Minckley, R.L., Packer, L. & Potts, S.G.
(2010) Ecological and life-history traits predict bee species responses to
environmental disturbances. Biological Conservation, 143, 2280–2291.
Williams, N.M. & Kremen, C. (2007) Resource distributions among habitats determine
solitary bee offspring production in a mosaic landscape. Ecological Applications,
17, 910–921.
Williams, N.M., Minckley, R.L. & Silveira, F.A. (2001) Variation in native bee faunas
and its implications for detecting community changes. Ecology and Society, 5, 7.
Williams, N.M., Ward, K.L., Pope, N., Isaacs, R., Wilson, J., May, E.A., et al. (2015)
Native wildflower plantings support wild bee abundance and diversity in
agricultural landscapes across the United States. Ecological Applications
(Accepted).
Williams, N.M. & Winfree, R. (2013) Local habitat characteristics but not landscape
urbanization drive pollinator visitation and native plant pollination in forest
remnants. Biological Conservation, 160, 10–18.
Wilson, J.S., Griswold, T. & Messinger, O.J. (2008) Sampling bee communities
(Hymenoptera: Apiformes) in a desert landscape: are pan traps sufficient? Journal
of the Kansas Entomological Society, 81, 288–300.
Winfree, R. (2010) The conservation and restoration of wild bees. Annals of the New
York Academy of Sciences, 1195, 169–197.

169

Winfree, R. (2013) Global change, biodiversity, and ecosystem services: What can we
learn from studies of pollination? Basic and Applied Ecology, 14, 453–460.
Winfree, R., Aguilar, R., Vazquez, D.P., LeBuhn, G. & Aizen, M.A. (2009) A metaanalysis of bees’ responses to anthropogenic disturbance. Ecology, 90, 2068–2076.
Winfree, R. & Kremen, C. (2009) Are ecosystem services stabilized by differences
among species? A test using crop pollination. Proceedings of the Royal Society BBiological Sciences, 276, 229–237.
Winfree, R., Williams, N.M., Dushoff, J. & Kremen, C. (2007) Native bees provide
insurance against ongoing honey bee losses. Ecology Letters, 10, 1105–13.
Wolf, A.T. & Ascher, J.S. (2008) Bees of Wisconsin (Hymenoptera: Apoidea:
Anthophila). The Great Lakes Entomologist, 41, 129–168.
Wood, T.J., Holland, J.M., Hughes, W.O.H. & Goulson, D. (2015) Targeted agrienvironment schemes significantly improve the population size of common
farmland bumblebee species. Molecular Ecology, early view. DOI:
10.1111/mec.13144.
Wright, I.R., Roberts, S.P.M. & Collins, B.E. (2015) Evidence of forage distance
limitations for small bees (Hymenoptera: Apidae). European Journal of
Entomology, 112, 1–8.
Wu, J.Y., Anelli, C.M. & Sheppard, W.S. (2011) Sub-lethal effects of pesticide residues
in brood comb on worker honey bee (Apis mellifera) development and longevity.
PLoS ONE, 6, e14720.
Zurbuchen, A., Cheesman, S., Klaiber, J., Muller, A., Hein, S. & Dorn, S. (2010a) Long
foraging distances impose high costs on offspring production in solitary bees.
Journal of Animal Ecology, 79, 674–681.
Zurbuchen, A., Landert, L., Klaiber, J., Muller, A., Hein, S. & Dorn, S. (2010b)
Maximum foraging ranges in solitary bees: only a few individuals have the
capability to cover long foraging distances. Biological Conservation, 143, 669–
676.

170