Draft - University of Toronto T-Space · Draft Climate and habitat influences on bee pollinator...

Draft

Climate and habitat influences on bee pollinator community

structure in Western Canada

Journal: Canadian Journal of Zoology

Manuscript ID cjz-2017-0226.R2

Manuscript Type: Article

Date Submitted by the Author: 31-Dec-2017

Complete List of Authors: Villalobos, Soraya; University of Calgary, Biological Science Vamosi, Jana C.; University of Calgary, Biological Science

Keyword: Pollination services, Environmental filtering, Phylogenetic diversity, Habitat specialization, Osmia, Andrena, Ceratina

https://mc06.manuscriptcentral.com/cjz-pubs

Canadian Journal of Zoology

Draft

1

Climate and habitat influences on bee community structure in Western Canada 1

2

Soraya Villalobos 3

Department of Biological Sciences, University of Calgary 4

2500 University Drive. N.W. 5

Calgary AB Canada 6

T2N-1N4 7

[email protected] 8

Phone: (+1) 4036164235 9

10

Jana C. Vamosi 11



Calgary AB Canada 14

T2N-1N4 15

[email protected] 16

Phone: +1 (403) 210-9594 17

18

19

Corresponding author 20

Soraya Villalobos 21



Calgary AB Canada 24

T2N-1N4 25

[email protected], [email protected] 26

Phone: (+1) 4036164235 27

28

29

30

31

32

33

34

35

Page 1 of 29



Draft

2

Abstract 36

The persistence of pollinators in a given habitat is determined in part by traits that affect 37

their response to environmental variables. Here, we show that climate and habitat 38

features are the main drivers of trait distribution in bees across spatially separated 39

habitats. We determined that trait and clade filtering results in bee assemblages in 40

Western Canada exhibiting clustering that is correlated with differences in temperature, 41

humidity and rainfall. Phylogenetic signals were detected in all traits associated with 42

pollinator life history strategies, including phenology. The Bombus (Latreille, 1802) clade 43

(including the social parasite subgenus Psithyrus Lepeletier, 1832) comprised a higher 44

proportion of prairie bees, whereas assemblages in Garry oak sites exhibited higher 45

representation from solitary bees (e.g., Osmia (Panzer, 1806); Andrena (Fabricius, 1775); 46

Ceratina, Latreille 1802). Because these same traits influence which plant species are 47

pollinated, this selective environmental occupancy within the two different habitats could 48

promote local adaptation of flowering plant species pollinated by more social clades in 49

prairies and more solitary bees in Garry oak. 50

51

Key words: Pollination services, habitat specialization, enviromental filtering, 52

phylogenetic diversity, Osmia, Andrena, Ceratina, Bombus, Psithyrus, mason bees, 53

mining bees, carpenter bees, bumble bees, cuckoo bumble bees 54

55

Introduction 56

57

Wild pollinators are key components of ecosystem function in terrestrial ecosystems 58

(Ollerton et al. 2011). Specifically, bees are considered the dominant provider of 59

Page 2 of 29



Draft

3

pollination services to wild plants and agricultural systems worldwide, supporting plant 60

reproduction and the maintenance of biodiversity (Davies et al. 2010). In North America, 61

native bees and honey bees contribute to the economy and the functioning of ecosystems, 62

with annual economic contributions estimated at ~$15 billion CDN (Kremen and Chaplin 63

2007). A worrisome trend is that bees are suffering severe population declines due to a 64

number of factors, including the use of pesticides, habitat loss, pests, diseases, invasive 65

species, and climate change (Potts et al. 2010). Current changes in land use can 66

negatively impact the availability of suitable habitats for nesting, shelter and feeding, and 67

impede their ability to shift ranges with climate (Kerr et al. 2015), yet the susceptibility 68

of different species of bees in Canada to changing environments is not currently 69

understood. 70

71

Pollinator composition within an assemblage is determined by three main factors: 1) 72

physiological tolerance limits to wind, temperature, and/or humidity; 2) floral resources; 73

and 3) nesting sites provided by habitat (Greenleaf et al. 2007; Zurbuchen et al. 2010; 74

Kennedy et al. 2013), yet the relative importance of these factors varies. In some 75

temperate ecosystems, air temperature appears as the main predictor of the activity of 76

pollinators overcoming the effect of flower abundance (Kühsel and Blüthgen 2015). 77

Interestingly, variation in thermal tolerances has also been related to body mass (Stone 78

and Willmer 1989). For example, smaller insects tend to be more active at lower 79

temperatures than larger species because higher temperatures can trigger greater water 80

loss (Chown and Terblance 2006). However, pollinator species do not respond equally to 81

temporal environmental changes (Frier et al. 2016). Wild bees (e.g., Bombus spp, Osmia 82

Page 3 of 29



Draft

4

cornuta Latreille, 1805) have demonstrated abilities in foraging under rainy and windy 83

conditions, unlike honey bees (Apis mellifera Linnaeus, 1758) whose activity 84

substantially decreases with the rain and high wind speed (Kennedy et al. 2013 Khüsel 85

and Blüthgen 2015). Determining the contribution of important functional traits, such as 86

body size, to the prevalence of genera in different environments would greatly increase 87

our abilities to predict which groups of pollinators are likely to be most impacted by 88

changing environments. 89

90

Flowering phenology and the presence of wild habitats are also critical for the persistence 91

of pollinators in fragmented ecosystems, such as urban-associated and agricultural 92

landscapes (Wray and Elle 2015). For instance, bee persistence is highly dependent of 93

abundant floral resources because: 1) the distance over which pollinators forage can 94

strongly affect their population dynamics; and 2) overlapping flowering phenologies in 95

species-rich plant communities might provide floral resources over a longer period of 96

time (Greenleaf et al. 2007). Likewise, the offer of specialized zygomorphic flowers, 97

which are predominant components of flowering resources in disturbed sites in Western 98

Canada, increases the chances of selecting for pollinators that can access the nectar of 99

these flowers (e.g., bees) (Villalobos et al. 2014). Change in vegetation also presents 100

variation in the quality and prevalence of nesting sites and therefore life history 101

characteristics, such as sociality and nesting preference, may also be associated with 102

pollinator occurrences in a given habitat (Zurbuchen et al. 2010). Similarly, habitat may 103

determine the mean size of pollinators in communities. For example, smaller bees have 104

shorter flight periods and experience earlier emergence, and may thus only appear in less 105

Page 4 of 29



Draft

5

disturbed environments (Greenleaf et al. 2007). Thus, colonization dynamics may result 106

in the appearance of habitat specialization driven by adaptations to specific 107

environmental factors (Kawecki and Ebert 2004; Fine and Kembel 2011). 108

Previous studies indicate that, within terrestrial Canadian ecozones, pollinator diversity is 109

influenced greatly by habitat characteristics, plant composition and climate (Bingham and 110

Orthner 1998; Sheffield et al. 2014; Straka and Starzomski 2015). For instance, prairie 111

sites in Alberta tend to support higher proportions flowering plant species with 112

morphological constraints (e.g., zygomorphy of Fabaceae (Villalobos and Vamosi 2016)) 113

compared to sites in the mixedwood plains ecozone and tall grass savannah in Ontario 114

and Quebec. These descriptive studies are important, but we have yet to uncover the 115

major determinants of broad-scale pollinator distribution differences across the different 116

ecozones of the country. 117

118

Understanding how pollinators vary in the quality of the pollination service to the plant 119

community is an emerging priority for ecosystem managers and is thought to be best 120

evaluated by measuring functional diversity (Sheffield et al. 2013). In particular, whether 121

the prevalence of certain traits varies according to tolerance to environmental factors is 122

an important first step in predicting whether pollination service to a plant community will 123

be adequate. In this study we aim to identify: 1) which pollinator lineages were 124

overrepresented in particular habitats; 2) what environmental variables contribute to the 125

assembly of local pollinator communities; and 3) which traits related with specialization 126

(e.g., solitary condition) are associated with a given environment. Understanding the 127

factors involved in structuring pollinator assemblages increases our capacity to predict 128

Page 5 of 29



Draft

6

what types of pollinators exhibit strict habitat requirements and will likely face increased 129

pressures with changes in the environment. 130

131

Methods 132

Local pollinator communities 133

We compiled a database of pollinator communities in coordination with the Canadian 134

Pollination Initiative (CANPOLIN), which collected data on potential plant visitors 135

throughout 2008-2012 across Canada (provided as a supplementary Table; Table S4). 136

CANPOLIN was a multidisciplinary research network designed to help characterize the 137

problem of pollinator decline in agricultural and natural ecosystems in Canada 138

(http://www.uoguelph.ca/canpolin/About/about.html), with an emphasis on bees. Here, 139

we analyzed the data collected in 35 sites from two different grassland habitats in 140

Western Canada: prairie (28 sites), and Garry oak savannah (seven sites) (CANPOLIN 141

2012) (Figure S1). Although Garry oak ecosystems could be categorized as a savannah, it 142

was maintained as a separated grassland habitat because of the complex arrangement of 143

trees (Quercus garryana Douglas ex Hook) and wildflowers 144

(http://www.goert.ca/about/index.php). Hymenopterans, specifically bees, are one of the 145

largest groups of insects that effectively pollinate flowering plants (Waser et al. 1996). 146

Though CANPOLIN involved a number of different collectors (see Table S4), the dataset 147

was most standardized in terms of collection protocols for bees collected from 2009-148

2011, excluding Apis mellifera. 149

150

Page 6 of 29



Draft

7

CANPOLIN carried out the data collection by setting out 30 coloured pans for a day 151

along a linear transect, and repeating this sampling every 7-10 days throughout the 152

season. Pans were placed ½- to ¾-full of soapy water (5 drops dish detergent per litre of 153

water) (http://www.uoguelph.ca/canpolin/Sampling/protocols.html). Ninety percent of 154

the taxa were identified to species level, while the remaining 10% were identified at the 155

genus or subfamilial level. 156

157

Traits and environmental variables 158

In compiling trait information, we used five traits that most likely influence community 159

assembly: phenology, life style, sociality, body size and nesting habit (Table 1, Table 1S). 160

This set of trait values was gathered from three different online databases, Scopus, Web of 161

Science, and Science Direct. The majority of trait values could be obtained by referring to 162

previously published data from Canadian ecozones, using the species and genus name as 163

key words. In addition, we further populated the trait dataset using information found in 164

books, articles and online taxonomic keys (e.g., http://bugguide.net/node/view/3/bgpage). 165

From these sources, we coded information on phenology, sociality, life style and nesting 166

habit as categorical traits (Table S1). Because cues for emergence are linked to climate, 167

phenology is a large determinate of pollinator assemblages collected at a certain point of 168

the season. To avoid circularity, the trait states found in the literature for phenology were 169

confirmed using CANPOLIN collection dates and the phenology records available in 170

https://www.inaturalist.org/. We then split the dataset for the species according to 171

whether the majority of the occurrences were in April-June (i.e., early season) or in July 172

to October (i.e., late season) (Table 1A) and analysed these subsets separately. The record 173

Page 7 of 29



Draft

8

of body size was categorized as both a continuous variable (i.e. body length) and a binary 174

variable (small <10mm and large >10 mm). Life style was coded as a binary trait 175

including parasitic and non-parasitic styles. Nesting habit was coded as a multistate 176

variable considering cavity/soil, rotten wood nesters, and cleptoparasites (Table 1B). 177

Cleptoparasitism complicates traits such as sociality because cleptoparasites, though not 178

social, tend to mimic their hosts. Sociality was therefore also categorized as a multistate 179

variable including eusocial, solitary and cleptoparasites species. We defined eusocial bees 180

as those having a wide range of cooperatives organizations forming large colonies (Brady 181

et al. 2006), solitary bees as those that work their nests with no co-operation among 182

individuals, although nests may be aggregated (Gathmann et al. 2016) and cleptoparasites 183

was defined as those species that do not build or provision their own nests; instead they 184

lay their eggs on the nest of other species (Rozen 2003). 185

186

We explored the variation in environmental conditions by performing a principal 187

components analysis (PCA). The biplot obtained with the analysis displayed the 188

distributions of sites in multivariate space defined by the set of three environmental 189

variables thought to heavily influence pollinator presence: temperature (°C), precipitation 190

(mm), and an indirect measure of humidity (heat degree days) (Figure 2). We further used 191

this information to assess the distribution of species abundances in relation to 192

environmental features in the RQL analysis. Environmental variables for collection dates 193

at each of the collection sites were obtained from Environment Canada weather stations. 194

(http://climate.weather.gc.ca/prods_servs/cdn_climate_summary_e. 195

html). 196

Page 8 of 29



Draft

9

197

Regional phylogeny 198

We generated a regional pollinator phylogeny using Phylomatic software (Webb and 199

Donoghue 2005), using an insect supertree with previously adjusted branch length, sensu 200

Chamberlain et al. (2014). To represent uncertainty in the topology (Cadotte et al. 2010), 201

the resolution of the supertree was standardized to genus level with polytomies linking 202

species within genus (Münkemüller et al. 2012). 203

204

Phylogenetic diversity 205

Phylogenetic diversity estimates were calculated with the picante package in R (Cowan et 206

al. 2016). We calculated the Faith’s phylogenetic diversity index (PD), and the Mean 207

Pairwise Distance (MPD) at each collection site weighted by species abundance. 208

Standardized effect sizes were calculated for MPD by comparing the observed 209

phylogenetic relatedness to a null model generated by using the option ‘phylogeny 210

shuffle’ that randomizes phylogenetic relationships among species (Webb et al. 2002). 211

212

Phylogenetic and taxonomic pollinator composition 213

To characterize pollinator beta-diversity we calculated Bray-Curtis distances between 214

sites (Bray and Curtis 1957). To assess phylogenetic dissimilarities of pollinator 215

composition among communities (PDB) we computed COMDIST, also referred as β-216

MPD (Baldeck et al. 2016). This function measures the mean phylogenetic distance 217

between pairs of species drawn from two spatially segregated communities (Kembel 218

2010). This metric captures the variation associated to internal nodes of phylogeny (e.g., 219

Page 9 of 29



Draft

10

genus and family level). Observed values of COMDIST were standardized with respect 220

to expected values calculated under phylogeny shuffle null model, which shuffles species 221

labels across the entire phylogeny (Phylocom V 4.1; Webb et al. 2011). Measures of 222

standardized effects (�NRI)werecalculatedasfollows: 223

224

βNRI = −1x(mean(�MPDrandom) − �MPDobserved)

!"(β#$%random)

225

Randomizations were repeated 99 times and the metric was calculated for abundance-226

weighted data. Thus, positives values of βNRI indicate that taxa are less closely related 227

between two samples while negative values indicate that taxa are more closely related 228

than expected by chance. 229

230

To test the extent to which habitat can explain variation in phylogenetic and taxonomic 231

beta-diversity we performed an ‘Adonis’ test using pairwise phylogenetic beta-diversity 232

matrices. In addition, we tested the overabundance of terminal nodes to determine which 233

clades contribute disproportionately to the phylogenetic structure using ‘Nodesig’ 234

function in Phylocom (Webb et al. 2011). Nodesig is an algorithm that uses a list of 235

locally coexisting species and evaluates for each clade on the regional phylogeny whether 236

the descendant species of a clade are present with higher or lesser frequency than 237

expected from the null model (Parra et al. 2010). 238

239

Phylogenetic signal 240

Page 10 of 29



Draft

11

The presence of phylogenetic signal for the binary traits of phenology, and life style was 241

assessed using the Fritz and Purvis test with 1,000 permutations (Fritz and Purvis 2010). 242

The test generates a value (D) that represents the magnitude of departure from both 243

random association and the clumping expected under a Brownian evolution threshold 244

model, using the R package “caper” (Orme et al. 2013). A D value below 0 reveals a 245

strongly clumped distribution, D � 0 means a “Brownian motion”- like evolutionary 246

distribution, D � 1 means random distribution, and D > 1 means overdispersed 247

distribution. 248

249

To evaluate the phylogenetic signal in quantitative and multi-state traits, i.e., body size, 250

sociality and type of nest, we used the root skewness test (Pavoine et al. 2011). This 251

analysis tests for concentrations of the trait in question on the phylogeny based on 252

permutations, testing trait states of a character either simultaneously or individually. 253

254

RLQ analysis 255

To identify statistically significant associations between environmental variables and 256

habitats (spatial variables), and clades and species traits, we applied RLQ analyses 257

(Pavoine et al. 2011). RLQ analysis is a multivariate ordination method used for 258

analyzing the main relationships between environmental gradients and species traits 259

mediated by species abundances (Pavoine et al. 2011; Peres-Neto and Kembel 2015). 260

Here we used the extended RLQ method combined with fourth corner analysis (Pavoine 261

et al. 2011), where traits and phylogeny of species are correlated with the geographic 262

locations and the environment where they occur. The method uses a phylogenetic 263

Page 11 of 29



Draft

12

eigenvector decomposition to express phylogenetic variation. The analysis starts with 264

four matrices: Matrix E for environmental variables, Matrix S for geographical space, 265

matrix T for biological traits and a matrix P for phylogenetic relationships among species. 266

A new matrix R (i.e., sites as rows and environmental and spatial coordinates as 267

columns) and Q (i.e., species as row and traits and phylogenetic variables as columns) 268

were linked to a matrix L, where rows are sites and columns are species and the entries 269

represent species abundance. The RLQ analysis uses a row permutation approach where 270

the two matrices R and Q are analysed with a multiple factorial method, and the matrix L 271

was analysed with correspondence analysis (Escofier and Pagès 1994). PQE and TQE 272

tests were used to assess whether there was phylogenetic and trait clustering across the 273

habitats, respectively (i.e., lower phylogenetic and trait diversity per site than expected 274

from the pool of species). Both PQE and TQE integrates distance among species and are 275

based on the entropy index (Pavoine et al. 2011) Likewise, we perform a Moran test 276

(Cliff and Ord 1973) to examine the geographical spatial autocorrelation in the dataset. 277

Although other approaches, mainly based on correlations and ordination techniques, have 278

been used to examine the association between traits, habitats and phylogenetic clades 279

(e.g., Beisner et al. 2010; Baldeck et al. 2016; Mayfield et al. 2010; Duarte et al. 2012), 280

we used the RLQ approach, because this method: 1) allows testing several traits, either 281

qualitative, binary or ordinal simultaneously; and 2) provides a broader description of 282

environmental filtering and traits in a phylogenetic and geographical context. This 283

method is available in Ade4 package in R (Dray et al. 2013). We used an R script adapted 284

from Pavoine et al. (2011). 285

286

Page 12 of 29



Draft

13

Results 287

The CANPOLIN pan-trap sampling method yielded 66 bee species in total from Western 288

Canadian sites. Twenty-four species were present in early summer samples, and 49 in 289

late summer samples (Table S1). Many of the species were solitary bees: we found 14 290

species of solitary bees in early summer and 10 solitary bee species in late summer 291

(Table S1, Figure 3). Furthermore, we detected five cleptoparasites species; all of them 292

present in early summer (April-June) (Table S1). 293

294

Phylogenetic diversity 295

In early season samples, Garry oak savannah sites had higher standardized phylogenetic 296

diversity than prairie measured as millions of years of evolutionary history (i.e., Mean 297

phylogenetic diversity = 763.4 my) compared to prairie (Mean phylogenetic diversity = 298

339.6 my) (Figure 2; t= 12.8, df=27 p-value < 0.05). In late season samples, we did not 299

detect significant differences between the two sites Garry oak (Mean phylogenetic 300

diversity= 948.3 my), and prairie sites (Mean phylogenetic diversity=871.8.4 my) (t = -301

0.69241, df = 9.09, p-value = 0.506). 302

303

However, in late season samples, mean pairwise distance (SES MPD) was significantly 304

phylogenetic clustered between the two habitats (p<0.05, Figure 1, Table S3), whereas 305

early season samples revealed a combination of clustering to randomness in phylogenetic 306

structure (p>0.05, Figure 1, Table S2). 307

308

Phylogenetic and taxonomic pollinator composition 309

Page 13 of 29



Draft

14

A Bray-Curtis dendrogram (Figure S2) showed that the two habitats were 310

phylogenetically clustered for different clades, with significant differences in species 311

composition (i.e., taxonomic composition is significantly dissimilar between them) 312

(Adonis analysis: F1, 33= 13.927; r2= 0.422; p<0.001). Likewise, the species composition 313

weighted by the evolutionary history of species (COMDIST distances, Figure S2), 314

revealed that the two habitats are dissimilar in their phylogenetic composition (Adonis 315

analysis: F1, 33= 5.1375; r2= 0.13471; p<0.001). The clades that significantly contributed 316

to the differences in pollinators’ assemblages were high Bombus species prevalence in the 317

prairies, and Osmia (Panzer, 1806), Andrena (Fabricius, 1775), Ceratina (Latreille 1802), 318

Stelis (Panzer, 1806) and Lasioglossum (Curtis, 1833) in Garry oak sites (Table 2). 319

320

Phylogenetic signal 321

We found strong phylogenetic signals for sociality, seasonality and body size in our 322

regional phylogeny (Table 1). For binary traits, D revealed a strongly clustered 323

distribution (Table 1A), significantly different from that observed in random 324

permutations. For multistate and quantitative traits retained in the permutation test, all 325

traits had significant phylogenetic signal (p<0.01) (Table 1B). 326

327

PCA and RLQ analysis 328

As expected, the PCA revealed contrasting environmental variables among sites (Figure 329

2). Further, it is important to note that all environmental variables exhibited spatial auto-330

correlation (Monte-Carlo Mantel test; Observation: -0.114; p=0.83), making the RLQ 331

analysis (which takes geographical position into account) a more conservative analysis. 332

Page 14 of 29



Draft

15

In the PCA in both early and late season, sites sampled in Garry oak were located in 333

zones with low humidity and high temperatures, while sites in prairies experienced more 334

precipitation and high values of humidity (Figure 2). 335

336

When compared across the two habitats, solitary condition and body length traits were 337

significantly correlated with environmental factors only in the late season (Monte-Carlo 338

Mantel test; Observation early season: 0.0045; p=0.669, Observation: 0.0169; p=0.755. 339

Monte-Carlo Mantel test; Observation late season: 0.213; p=0.001, Observation: 0.158; 340

p=0.004). 341

For the pool of species collected in the early season, the TQE test revealed trait 342

overdispersion across the two western habitats in early season (Monte-Carlo Mantel test; 343

Observation: 0.161; p=0.483) and trait clustering in late season (Monte-Carlo Mantel test; 344

Observation: 0.266; p=0.006). The PQE function showed significant phylogenetic 345

clustering in all of the habitats in late season (Mantel test; Observation: 0.166; p=0.049), 346

and random structure in early season (Monte-Carlo Mantel test; Observation: 0.132; 347

p=0.579). This result corroborates the results found with SES MPD analysis described 348

above. 349

350

The first axis of RLQ applied to both space and environment, and both traits and 351

phylogeny explained 48.3% of the total variation for early season, while for late season 352

the data explained 79.9% of the variation. This means that for late season there was a 353

significant overall relationship between environmental variables and species traits. In the 354

ordination plot of traits along the first RLQ axis (Figure S3), the negative side of this axis 355

Page 15 of 29



Draft

16

correspond to areas that exhibited more precipitation (i.e., prairies), whereas the positive 356

side of the first axis of the RLQ analysis were areas that exhibited high temperatures, 357

(i.e., Garry oak meadows) 358

359

The trait associated with areas that exhibited more precipitation was soil nesting (Figure 360

S3). Traits related to areas that exhibited high temperatures were cavity nesting, and 361

cleptoparasitism (Figure S3). 362

363

DISCUSSION 364

365

Within Western Canada, we find that bees show considerable turnover between habitats. 366

We also find that bee assemblages are aligned with climate and habitat features; 367

environmental conditions explained 48% of the total variation in the early season and 368

79% of the total variation in species composition in the late season. Specifically traits that 369

foster persistence in certain climates and habitats (e.g., nesting habits, body size) appear 370

to evolve relatively slowly such that the signature of common ancestry is retained and we 371

can thus observe differences in the clades of insects collected among different habitats. 372

Consequently, environmental filters can contribute to the phylogenetic structuring of 373

local bee communities. 374

375

The degree of spatial structure (i.e., spatial autocorrelation), environmental filtering and 376

the mobility of species jointly influence the distribution of species traits (Cadotte et al. 377

2010). We posit that the differential occupancy in the two different habitats was due in 378

Page 16 of 29



Draft

17

part to the different environmental tolerances of certain pollinator clades (Weigelt et al. 379

2015). For instance, the clade that significantly contributed to the community diversity in 380

prairies was the social bee genus Bombus (Table 2). Bombus species can forage under 381

humid and rainy weather conditions, possibly as an adaptive strategy to lessen 382

competition with more aggressive species (Frier et al. 2016; Keppner and Jarau 2016). 383

Similarly, small-bodied bees with metallic colors were the more important group in 384

warmer sites (e.g., Garry oak), where the reflection of solar radiation is critical to better 385

control overheating. This likely corroborates the idea that small sizes and reflecting 386

colors in insects are adaptations to tolerate heat (Kühsel and Blüthgen 2015). 387

388

While nesting habits of pollinators may not immediately be perceived as related to 389

climate, recent findings indicate that they are critically important in regulating their 390

tolerance of climate variability (Kühsel and Blüthgen 2015). We detected that the cavity 391

nesters and cleptoparasitism traits were positively correlated with habitat where 392

temperature was the strongest determinant of species composition (i.e., Garry oak; Figure 393

S3). These differences in nesting habits in different habitats may reflect how cavity 394

nesting can dampen microclimatic changes, especially late in the season when the 395

temperature substantially increases (Amat-Valero et al. 2013). The local association of 396

cleptoparasites showed a strong phylogenetic signal (Table 1A), and was well explained 397

by the phylogenetic clustering of the ecological traits filtered by the environment, 398

specifically for late season samples (Table S1, S3). Taken together, these findings might 399

suggest that cleptoparasitism on certain clades of solitary bees could be more specialized 400

than previously thought. Further studies should be done to determine to which extent the 401

Page 17 of 29



Draft

18

presence of cleptoparasites is influenced by the presence of cavity and soil nesters hosts. 402

Flowering plant resources in the Garry oak-savannah tend to influence the assemblage of 403

bee composition that, in turn, effectively pollinate plants in oak-savannah fragments 404

(Fuchs et al. 2001). Although CANPOLIN did not measure foraging sources within Garry 405

oak fragments, we posit that the overrepresentation of small solitary bees (e.g., Andrena, 406

Ceratina, Osmia), which are arguably more specialized (Macivor et al. 2014), and some 407

dietary specialist clades (e.g., the Asteraceae specialist Andrena) may indicate the driving 408

force behind floral evolution in these areas (e.g., if small bees drive the parallel evolution 409

of small flowers). Furthermore, this overrepresentation based on the analysis of 410

phylogeny and species richness (Nodesig analysis, Table 2) suggests that these clades are 411

responsible for the patterns of phylogenetic structure detected in Garry oak communities 412

(Parra et al. 2010; Adderley and Vamosi 2015). However, we still require studies on 413

foraging preferences and pollination effectiveness to have a more complete understanding 414

of the role of solitary bees in this habitat. 415

416

We chose traits that are considered important in describing life history strategies in 417

seasonal environments (Table S1) and indeed trait distribution exhibited clear patterns of 418

distribution along gradients in temperature, precipitation, and humidity. These same traits 419

exhibit phylogenetic inertia such that the features of the environment that characterise 420

each habitat may suggest stabilizing selection and local adaptation (Ravigne et al. 2009), 421

specifically in the social clades of prairies in the solitary bee clades found in Garry oak. 422

Local adaptive response has long been recognized as a critical aspect of ecological 423

specialization (Kawecki and Ebert 2004), and the overrepresentation of the clades that we 424

Page 18 of 29



Draft

19

detected may be a result of in situ diversification of lineages in their respective favoured 425

habitats (Parra et al. 2010). In other words, environmental filtering may lead to the stable 426

coexistence of locally specialized species (Weigelt et al. 2015), which might facilitate the 427

evolution of further specialization in these clades. 428

429

CONCLUSION 430

The effect of environmental filters on the assemblage of the local bee communities 431

reinforces the idea that phylogenetic constraints account for habitat specialization. Our 432

findings suggest that the prevalence of particular clades of pollinators in the two different 433

habitats is partly due to the environmental tolerances of certain pollinator clades. We 434

hypothesize that features of the environment that characterise each habitat, as well as the 435

greater degree of phylogenetic clustering in floral visitors, will promote the local 436

adaptation of the social genus Bombus in prairies and more solitary clades in Garry oak 437

savannah. Because solitary clades are considered more specialized and more often 438

restricted in which flowering plant species they can access, Garry oak communities may 439

require more conservation efforts to maintain adequate pollination services to plant 440

communities. 441

442

ACKNOWLEDGEMENTS 443

This is publication no. 148 of CANPOLIN 444

445

446

Page 19 of 29



Draft

20

REFERENCES 447 448 Amat-Valero, M., Calero-Torralbo, M. A, and Valera, F. 2013. Temperature during the 449

free-living phase of an ectoparasite influences the emergence pattern of the infective 450 phase. Parasitology, 140(11): 1357–67. doi:10.1017/S0031182013000929. 451

Baldeck, C.A., Harms, S.W.K.K.E., and John, J.B.Y.R. 2016. Phylogenetic turnover 452 along local environmental gradients in tropical forest communities. Oecologia, 453 182(2): 547–557. Springer Berlin Heidelberg. doi:10.1007/s00442-016-3686-2. 454

Beisner, B.E., Peres-Neto, P.R., Lindström, E.S., Barnett, A., Longhi, L., and Beisner, E. 455 2010. The Role of Environmental and Spatial Processes in Structuring Lake 456 Communities from Bacteria to Fish. Ecological Society of America, 87(12): 2985–457 2991. 458

Bingham, R.A., and Orthner A.R. 1998. Efficient pollination of alpine plants. Nature, 459 391(6664): 238-239. http://dx.doi.org/10.1038/34564. 460

Bray, J.R., and Curtis, J.T. 1957. An Ordination of the upland forest community of 461 southern Wisconsin. Ecol. Monogr. 27: 325-349. doi:10.2307/1942268. 462

Brady, G., Sipes, S., Pearson, A., and Danforth, B.N. 2006. Recent and simultaneous 463 origins of eusociality in halictid bees. Proc Biol Sci. 273: 1643–1649. 464 doi:10.1098/rspb.2006.3496. 465

Cadotte, M.W., Jonathan Davies, T., Regetz, J., Kembel, S.W., Cleland, E., and Oakley, 466 T.H. 2010. Phylogenetic diversity metrics for ecological communities: integrating 467 species richness, abundance and evolutionary history. Ecol. Lett. 13(1): 96–105. 468 doi:10.1111/j.1461-0248.2009.01405.x. 469

CANPOLIN. 2012. The Canadian Pollination Initiative. Available from: 470 http://www.uoguelph.ca/canpolin/Publications/CANPOLIN%20background.pdf 471

Chamberlain, S., Cartar, R.V.,Worley, A.C., Semmler, S.J., Gielens, G., Elwell, S., 472 Evans, M.E., Vamosi, J.C., Elle, E. 2014. Traits and phylogenetic history contribute 473 to network structure across Canadian plant-pollinator communities. Oecologia, 474 176:545-56. 475

Chown, S., and Terblance, J.S., 2006. Physiological Diversity in Insects: Ecological and 476 Evolutionary Contexts. Adv. Insect Phys. 33: 50–152. doi:10.1016/S0065-477 2806(06)33002-0. 478

Cliff, A.D., and Ord, J.K. 1973. Spatial Autocorrelation. Pion. London. 479 Cowan, P.D., Hel-, M.R., Morlon, H., Webb, O., and Kembel, M.S.W. 2016. Package 480

“picante”. available from: https://cran.r-481 project.org/web/packages/picante/picante.pdf 482

Davies, K.F., Melbourne, B. A., James, C.D., and Cunningham, R.B. 2010. Using traits 483 of species to understand responses to land use change: Birds and livestock grazing in 484 the Australian arid zone. Biol. Conserv. 143(1): 78–85. 485 doi:10.1016/j.biocon.2009.09.006. 486

Dray, S., Dufour, A.-B., Thioulouse, J., Jombart, T., Pavoine, S., Lobry, J.R., Ollier, S., 487 and Siberchicot, A. 2013. Package “ade4.” : 1–369. Available from http://pbil.univ-488 lyon1.fr/ADE-4. 489

Duarte, L.D.S., Prieto, P. V, and Pillar, V.D. 2012. Assessing spatial and environmental 490 drivers of phylogenetic structure in Brazilian Araucaria forests. Ecography 35 (10): 491 952–960. doi:10.1111/j.1600-0587.2011.07193.x. 492

Page 20 of 29



Draft

21

Escofier, B., and Pagès, J. 1994. Multiple factor analysis (AFMULT package). Comput. 493 Stat. Data Anal. 18(1): 121–140. doi:10.1016/0167-9473(94)90135-X. 494

Fine, P.V.A., and Kembel, S.W. 2011. Phylogenetic community structure and 495 phylogenetic turnover across space and edaphic gradients in western Amazonian tree 496 communities. Ecography 34(4): 552–565. doi:10.1111/j.1600-0587.2010.06548.x. 497

Frier, S.D., Somers, C.M., Sheffield C. 2016. Comparing the performance of native and 498 managed pollinators of Haskap (Lonicera caerulea: Caprifoliaceae), an emerging 499 fruit crop. Agriculture, Ecosystems and Environment, 219: 42-48. 500

Fritz, S.A., and Purvis, A. 2010. Selectivity in mammalian extinction risk and threat 501 types: A new measure of phylogenetic signal strength in binary traits. Conserv. Biol. 502 24(4): 1042–1051. doi:10.1111/j.1523-1739.2010.01455.x. 503

Fuchs, M.A., Bio, R.P., Street, M., VS, B.C., Marsha, I., Cannings, S., Ceska, A., 504 Costanzo, B., Groot, D., Douglas, G., Duncan, B., Erickson, W., Fairbarns, M., 505 Fleming, T., Flynn, S., Fraser, D., Gillespie, T., Harcombe, G., Illingworth, J., 506 Krannitz, P., Maslovat, C., Morrison, K., Penny, J., Roemer, H., Shepard, M., 507 Spencer, F., Erickson, W., and Spencer, F. 2001. Towards a Recovery Strategy for 508 Garry Oak and Associated Ecosystems in Canada : Ecological Assessment and 509 Literature Review. Available from http://www.goert.ca/documents/doc-Eco-510 Assessment-Lit-Review-2001.pdf 511

Gathmann, A., Tscharntke, T. 2016. Foraging Ranges of Solitary Bees. British Ecological 512 Society, 71(5): 757–764. 513

Greenleaf, S.S., Williams, N.M., Winfree, R., Kremen, C. 2007. Bee foraging ranges and 514 their relationship to body size. Oecologia 153(3): 589–596. doi:10.1007/s00442-007. 515

Kawecki, T.J., and Ebert, D. 2004. Conceptual issues in local adaptation. Ecol. Lett. 516 7(12): 1225–1241. doi:10.1111/j.1461-0248.2004.00684.x. 517

Kembel, S. 2010. An introduction to the picante package. R Proj. Available from 518 http://picante.r-forge.r-project.org/picante-intro.pdf 519 doi:10.1093/bioinformatics/btq166. 520

Kennedy, C.M., Lonsdorf, E., Neel, M.C., Williams, N.M., Taylor, H., Winfree, R., 521 Brittain, C., Alana, L., and Cariveau, D. 2013. A global quantitative synthesis of 522 local and landscape effects on wild bee pollinators in agroecosystems. Ecology 523 Letters, 16(5): 584–599. doi:10.1111/ele.12082. 524

Keppner, E.M., and Jarau, S. 2016. Influence of climatic factors on the flight activity of 525 the stingless bee Partamona orizabaensis and its competition behavior at food 526 sources. Behav. Physiol. 202(9–10): 691–699. doi:10.1007/s00359-016-1112-1. 527

Kerr, J.T., Pindar, A., Galpern, P., Packer, L., Potts, S.G., Roberts, S.M., Rasmont, P., 528

Schweiger, O., Colla, S.R., Richardson, L.L., Wagner, D.L., and Gall, L.F. 2015. 529

across continents. Science, 349(6244): 177–180. 530 Kremen, C. and Chaplin, R. 2007. Insects as providers of ecosystem services: crop 531

pollination and pest control. In Insect conservation biology. Proceedings of the 532 Royal Entomological Society's 23rd Symp. Edited by. A. J. A. Stewart, T. R. New, 533 and O. T. Lewis. Wallingford, UK: CABI Publishing. pp 203-218. 534

Kühsel, S., and Blüthgen, N. 2015. High diversity stabilizes the thermal resilience of 535 pollinator communities in intensively managed grasslands. Nature Communications 536 6 (7989). doi:10.1038/ncomms8989. 537

Macivor, J.S., Cabral, J.M., and Packer, L. 2014. Pollen specialization by solitary bees 538

Page 21 of 29



Draft

22

in an urban landscape. (June 2013): 139–147. doi:10.1007/s11252-013-0321-539

4. 540 Mayfield, M.M., Bonser, S.P., Morgan, J.W., Aubin, I., McNamara, S., and Vesk, P. A. 541

2010. What does species richness tell us about functional trait diversity? Predictions 542 and evidence for responses of species and functional trait diversity to land-use 543 change. Glob. Ecol. Biogeogr. 19 (4): 423–431. doi:10.1111/j.1466-544 8238.2010.00532.x. 545

Münkemüller, T., Lavergne, S., Bzeznik, B., Dray, S., Jombart, T., Schiffers, K., and 546 Thuiller, W. 2012. How to measure and test phylogenetic signal. Methods Ecol. 547 Evol. 3(4): 743–756. doi:10.1111/j.2041-210X.2012.00196.x. 548

Ollerton, J., Winfree, R., and Tarrant, S. 2011. How many flowering plants are pollinated 549 by animals. Oikos, 120(3): 321–326. doi:10.1111/j.1600-0706.2010.18644.x. 550

Orme, D., Freckleton, R., Thomas, G., Petzoldt, T., Fritz, S., Isaac, N., and Pearse, W. 551 2013. Caper: Comparative Analyses of Phylogenetics and Evolution in R. R package 552 (version 0.5.2). (Vienna, R Foundation for Statistical Computing.). Available from 553 https://cran.r-project.org/web/packages/caper/caper.pdf [Accesed November 2016]. 554

Parra, J.L., McGuire, J. A, and Graham, C.H. 2010. Incorporating clade identity in 555 analyses of phylogenetic community structure: an example with hummingbirds. Am. 556 Nat. 176(5): 573–87. doi:10.1086/656619. 557

Pavoine, S., Vela, E., Gachet, S., De Bélair, G., and Bonsall, M.B. 2011. Linking patterns 558 in phylogeny, traits, abiotic variables and space: A novel approach to linking 559 environmental filtering and plant community assembly. [online]. J. Ecol. 99(1): 560 165–175. doi:10.1111/j.1365-2745.2010.01743.x. 561

Peres-Neto, P.R., and Kembel, S.W. 2015. Phylogenetic gradient analysis: environmental 562 drivers of phylogenetic variation across ecological communities. Plant Ecology, 563 216:709-724. 564

Potts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger, O., and Kunin, W.E. 565 2010. Global pollinator declines: Trends, impacts and drivers. Trends Ecol. Evol. 566 25(6): 345–353. doi:10.1016/j.tree.2010.01.007. 567

Ravigne, V., Dieckmann, U., and Olivieri, I. 2009. Live Where You Thrive: Joint 568 Evolution of Habitat Choice and Local Adaptation Facilitates Specialization and 569 Promotes Diversity. Am. Nat. 174(4): E141–E169. doi:10.1086/605369. 570

Rozen, J.G. 2003. Eggs , Ovariole Numbers , and Modes of Parasitism of Cleptoparasitic 571 Bees , with Emphasis on Neotropical Species ( Hymenoptera : Apoidea ). Am. mus. 572 novit 3413: 1–36. 573

Sheffield, C. S., S. D. Frier, and S. Dumesh. 2014. The Bees (Hymenoptera : Apoidea , 574 Apiformes ) of the Prairies Ecozone with Comparisons to other Grasslands of 575 Canada. In Arthropods of Canadian Grasslands (Volume 4): Biodiversity and 576 Systematics Part 2. Edited by D. J. Giberson and H. A. Cárcamo. Biological Survey 577 of Canada. pp. 427-467. © 2014 Biological Survey of Canada. [online]. ISBN 978-578 0-9689321-7-9 doi: http://dx.doi.org/10.3752/9780968932179.ch11 579

Sheffield, C.S., Pindar, A., Packer, L., and Kevan, P.G. 2013. The potential of 580 cleptoparasitic bees as indicator taxa for assessing bee communities. Apidologie 581 44(5): 501–510. doi:10.1007/s13592-013-0200-2. 582

Stone, B.Y.G.N., and Willmer, P.G. 1989. Warm-Up Rates And Body Temperatures In 583 Bees : The Importance Of Body Size , Thermal Regime And Phylogeny. Journal of 584

Page 22 of 29



Draft

23

Experimental Biology, 328: 303–328. 585 Straka, J.R., and Starzomski, B.M. 2015. Fruitful factors: what limits seed production of 586

flowering plants in the alpine? Oecologia, 178(1): 249–260. doi:10.1007/s00442-587 014-3169-2. 588

Villalobos, C., González-Carcacía, J.A., Rodríguez, J.P., and Nassar, J. 2014. 589

Interspecific and Interannual Variation in Foliar Phenological Patterns in a 590

Successional Mosaic of a Dry Forest in the Central Llanos of Venezuela. Taylor 591

and Francis Group, Boca Raton, Florida. 592 Villalobos, S., and Vamosi, J.C. 2016. Increasing land use drives changes in plant 593

phylogenetic diversity and prevalence of specialists. PeerJ, 4: e1740. 594 doi:10.7717/peerj.1740. 595

Waser, N. M., Chittka, L., Price, M. V., Williams, N. M., and Ollerton, J.1996. 596 Generalization in pollination systems, and why it matters. Ecology 77: 1043–1060. 597

Webb, C., Ackerly, D., and Kembel, S. 2011. Phylocom: Software for the analysis of 598 phylogenetic community structure and character evolution (with phylomatic and 599 ecovolve). Phylocom Man. Available from 600 file:///C:/Users/Steph_Admin/Desktop/Chapter 4/phylocom_manual.pdf. 601

Webb, C.O., Ackerly, D.D., McPeek, M. A., and Donoghue, M.J. 2002. Phylogenies and 602 Community Ecology. Annu. Rev. Ecol. Syst. 33(1): 475–505. 603 doi:10.1146/annurev.ecolsys.33.010802.150448. 604

Webb, C.O., and Donoghue, M.J. 2005. Phylomatic: tree assembly for applied 605 phylogenetics. Mol. Ecol. Notes. 5(1): 181–183. doi:10.1111/j.1471-606 8286.2004.00829.x. 607

Weigelt, P., Daniel Kissling, W., Kisel, Y., Fritz, S. A., Karger, D.N., Kessler, M., 608 Lehtonen, S., Svenning, J.-C., and Kreft, H. 2015. Global patterns and drivers of 609 phylogenetic structure in island floras. Sci. Rep. 5: 12213. doi:10.1038/srep12213. 610

Wray, J.C., and Elle, E. 2015. Flowering phenology and nesting resources influence 611 pollinator community composition in a fragmented ecosystem. Landscape Ecology 612 30: 261–272. doi:10.1007/s10980-014-0121-0. 613

Zurbuchen, A., Landert, L., Klaiber, J., Müller, A., Hein, S., and Dorn, S. 2010. 614 Maximum foraging ranges in solitary bees : only few individuals have the capability 615 to cover long foraging distances. Biol. Conserv. 143(3): 669–676. 616 doi:10.1016/j.biocon.2009.12.003. 617

618 619 620 621 622 623 624 625 626 627 628 629 630

Page 23 of 29



Draft

24

631 632 633 634 635 636 637 638 639 640

FIGURES CAPTIONS 641 642 Figure 1. Measures of: A) phylogenetic diversity (PD) and B) Standardized Mean 643 Pairwise Distance (SES MPD) for two habitats, prairies and Garry oak. The habitats have 644 significant phylogenetic clustering in species when phylogenetic diversity was compared 645 to that expected from a null phylogeny, SES MPD (P < 0.05) and random structure SES 646 MPD (P>0.05), for both early and late summer season. 647 648 Figure 2. Principal component analysis (PCA) based in three environmental variables of 649 three habitats of western Canada. Temp C= temperature °C, PP.mm= precipitation, and 650 Humidity for A) early summer season B) late summer season. 651 652 Figure 3. Regional bee phylogeny showing the clades present in early and late season. 653 Highlighted clades correspond to species of the important clades in Prairies (green band) 654 and Garry oak meadows (red band). 655 656

Page 24 of 29



Draft0

1000

2000

3000

Garryoak Prairie

PD

A

-4

-2

0

2

Garryoak Prairie

MPD

C

0

1000

2000

3000

Garryoak Prairie

PD

B

-4

-2

0

2

Garryoak Prairie

MPD

D

Page 25 of 29



DraftTemp.C

Humidity

PP.m

m

-3

-2

-1

0

1

2

-2 -1 0 1 2 3

standardized PC1 (76.3% explained var.)

sta

nd

ard

ize

d P

C2

(2

3.7

% e

xp

lain

ed

va

r.)

garryoak prairie

A

Temp.C

Humidity

PP.mm

-3

-2

-1

0

1

2

-2 -1 0 1 2 3

standardized PC1 (89.3% explained var.)

sta

nd

ard

ize

d P

C2

(1

0.4

% e

xp

lain

ed

va

r.)

garryoak prairie

B

Page 26 of 29



Draft

Early season Late season

Garry oak meadows

Prairies

Page 27 of 29



Draft

TABLES

Table 1. A) Phylogenetic signal in pollinator binary traits.

Trait Type Estimated D PE(D)

random*

PE(D)

Brownian** Conclusion

Phenology

(seasonality)

binary 0.302729 0 0.718 Signal

Sociality (solitary vs.

social)

binary -0.712703 0 0.967 Signal

Life style (parasitic

vs. free)

binary -0.638517 0 0.839 Signal

Body size (small vs.

large)

binary -0.037333 0 0.568 Signal-low

*Probability of E(D) resulting from no (random) phylogenetic structure

**Probability of E(D) resulting from Brownian phylogenetic structure

Table 1 B) Phylogenetic signal in nest type, sociality and body length.

Trait

Type

Deviation from

theoretical values P-value

Conclusion

Nest type

(traits considered together:

nesting cavity, soil, rotten

wood, cleptoparasite)

Multistate

0.3051737 P<0.01 Signal

Sociality

solitary, eusocial, cleptoparasite

Multistate

0.2260277 P<0.001 Signal

Body length Quantitative 0.4112965 P<0.01 Signal

Page 28 of 29



Draft

Table 2. Clades significantly contributing to the pollinator community structure in two

different habitats of western Canada (More column: overrepresented, Less column:

underrepresented). Nodesig algorithm (Phylocom).

Habitat More Less

Lassioglossum

Halictus

Stelis

Osmia

Apidae

Garry oak Andrena

Ceratina

Prairies Bombus Osmia

Megachile

Page 29 of 29



Draft - University of Toronto T-Space · Draft Climate and habitat influences on bee pollinator...

Documents

Transcript of Draft - University of Toronto T-Space · Draft Climate and habitat influences on bee pollinator...