University of Zurich · predictions of increasing specialization of pollination systems during succession, but are in agreement with expectations from optimal foraging theory, predicting

University of ZurichZurich Open Repository and Archive

Winterthurerstr. 190

CH-8057 Zurich

http://www.zora.uzh.ch

Year: 2010

Plant-pollinator network assembly along the chronosequence of aglacier foreland

Albrecht, M; Riesen, M; Schmid, B

Albrecht, M; Riesen, M; Schmid, B (2010). Plant-pollinator network assembly along the chronosequence of aglacier foreland. Oikos, 119(10):1610-1624.Postprint available at:http://www.zora.uzh.ch

Posted at the Zurich Open Repository and Archive, University of Zurich.http://www.zora.uzh.ch

Originally published at:Oikos 2010, 119(10):1610-1624.

Albrecht, M; Riesen, M; Schmid, B (2010). Plant-pollinator network assembly along the chronosequence of aglacier foreland. Oikos, 119(10):1610-1624.Postprint available at:http://www.zora.uzh.ch

Posted at the Zurich Open Repository and Archive, University of Zurich.http://www.zora.uzh.ch

Originally published at:Oikos 2010, 119(10):1610-1624.

Plant-pollinator network assembly along the chronosequence of aglacier foreland

Abstract

Forelands of retreating glaciers offer an ideal model system to study community assembly processesduring primary succession. As plants colonize the area that is freed from ice they should beaccompanied by their pollinators to successfully reproduce and spread. However, little is known aboutthe assembly of plant-pollinator networks. We therefore used quantitative network analysis to study thestructure of plant-pollinator interactions at seven sites representing a chronosequence from 8 to 130years since deglaciation on the foreland of the Morteratsch glacier (southeastern Switzerland). At thesesites, individual visits of plant fl owers by insects were recorded throughout the fl owering season.Species richness of insect-pollinated plants and plant-pollinating insects, together with measures ofinteraction diversity and evenness, increased along the chronosequence at least for the fi rst 80 yearsafter deglaciation. Bees were the most frequent fl ower visitors at the two youngest sites, whereas fl iesdominated in mature communities. Pollinator generalization (the number of visited plant speciesweighted by interaction strength), but not plant generalization, strongly increased during the primarysuccession. This was reflected in a pronounced decline in network level specialization (measured asBlüthgen's H2') and interaction strength asymmetry during the fi rst 60 years along the chronosequence,while nestedness increased along the chronosequence. Thus, our findings contradict niche-theoreticalpredictions of increasing specialization of pollination systems during succession, but are in agreementwith expectations from optimal foraging theory, predicting an increase in pollinator generalization withhigher plant diversity but similar flower abundance, and an increase in diet breadth at higher pollinatordensities during primary succession.

Albrecht et al. Plant–pollinator network assembly 1 _______________________________________________________________________

1 2 3 4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

Plant–pollinator network assembly along the chronosequence of a glacier foreland

Matthias Albrecht 1, Matthias Riesen 1 and Bernhard Schmid 1

1 Institute of Evolutionary Biology and Environmental Studies, University of Zurich,

Winterthurerstrasse 190, 8057 Zurich, Switzerland

Corresponding author:

Matthias Albrecht

E-mail: [email protected]

Fax: +41 44 635 57 11


Abstract 22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

Forelands of retreating glaciers offer an ideal model system to study community

assembly processes during primary succession. As plants colonize the area that is

freed from ice they should be accompanied by their pollinators to successfully

reproduce and spread. However, little is known about the assembly of plant–

pollinator networks. We therefore used quantitative network analysis to study the

structure of plant–pollinator interactions at seven sites representing a chronosequence

from 8 to 130 years since deglaciation on the foreland of the Morteratsch glacier (SE-

Switzerland). At these sites, individual visits of plant flowers by insects were

recorded throughout the flowering season. Species richness and Shannon diversity of

insect-pollinated plants and plant-pollinating insects, together with measures of

interaction diversity and evenness, increased along the chronosequence at least for the

first 80 years after deglaciation. Bees were the most frequent flower visitors at the

two youngest sites, whereas flies dominated in mature communities. Pollinator

generalization (the number of visited plant species weighted by interaction strength),

but not plant generalization, strongly increased during the primary succession. This

was reflected in a pronounced decline in network level specialization (measured as

Blüthgen’s H2’) and interaction strength asymmetry during the first 60 years along the

chronosequence, while nestedness increased along the chronosequence. Thus, our

findings contradict niche-theoretical predictions of increasing specialization of

pollination systems during succession, but are in agreement with expectations from

optimal foraging theory, predicting an increase in pollinator generalization with

higher plant diversity but similar flower abundance, and an increase in diet breadth at

higher pollinator densities during primary succession.


Introduction 46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

How communities assemble is a central issue in ecology (e.g. Clements 1916, Hubell

2001, Lortie et al. 2004, Sargent and Ackerly 2008). Primary successions, in

particular glacier forelands with a well-known chronology of glacier retreats, have a

long history as natural model systems for the study of plant community assembly

(reviewed in Matthews 1992; Walker and del Moral 2003). The chronosequence

across a glacier foreland can be interpreted, with the required caution, as a space for

time substitution of primary succession (Foster and Tilman 2000, Kaufmann 2001,

Walker and del Moral 2003). Previous studies have focused on a single trophic level,

typically plant (Matthews 1992) and less often animal community assembly

(Kaufmann 2001, Hodkinson et al. 2001, Hodkinson et al. 2004). Very little is known

about multitrophic community assembly and how plant communities interact with

animal communities during this process (Sargent and Ackerly 2008). Here, we study

a plant–pollinator community assembly process during primary succession on a

glacier foreland.

Most of the research on plant–pollinator interactions has tended to focus on

single species or guilds of related plants and their pollinators, giving little attention to

the larger community context (Waser et al. 1996). Recently, however, plant–

pollinator interactions at the community level have attracted considerable attention

(Waser and Ollerton 2006 and references therein), largely fueled by new

developments in the analysis of complex networks (Proulx et al. 2005). Complex

network analysis has provided important insight into the structure of plant–pollinator

networks (e.g. Jordano 1987, Olesen and Jordano 2002, Jordano et al. 2003, Stang et

al. 2006, Bascompte and Jordano 2007, Blüthgen et al. 2008, Ings et al. 2009,

Vasquez et al. 2009a,b). For example, an increasing number of studies suggest that


the structure of most plant–pollinator networks is asymmetric (Vazquez and Aizen

2004, Bascompte et al. 2006) and nested (Bascompte et al. 2003), whereby

specialized species interact with a subset of the interaction partners of more

generalized species. Furthermore, large interaction networks may show some degree

of modularity (Olesen et al. 2007). Several ecological and evolutionary determinants

of these network patterns are currently discussed. The concept of forbidden links (e.g.

Jordano et al. 2003) refers to the possibility that some interactions may not occur

because of mismatches in the morphology, phenology or spatial distribution of

species (Stang et al. 2006, Bascompte and Jordano 2007, Santamaría and Rodríguez-

Gironés 2007). Conversely, the neutrality hypothesis claims that network patterns are

mainly driven by random encounters, so that abundant species interact more

frequently and with a higher number of species than rare species (Vazquez 2005).

Recent findings suggest that both of these mechanisms contribute to observed plant–

pollinator network structures (Blüthgen et al. 2006, Stang et al. 2006, Bascompte and

Jordano 2007, Vazquez et al. 2009a,b).

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

Much less is known about the dynamics of plant–pollinator networks (but see

e.g. Petanidou et al. 2008, Olesen et al. 2008, Bascompte and Stouffer 2009). A

fundamental open question is how network structure changes during the process of

network assembly (Olesen et al. 2008, Bascompte and Stouffer 2009). Some studies

used simulation models to predict network disassembly regarding robustness and

stability (see Bascompte and Jordano 2007, Bascompte and Stouffer 2009 and

references therein), concluding that the heterogeneous degree distribution and the

cohesive organization in nested systems, as typically found for mutualistic networks,

may primarily account for the high robustness predicted for networks that are

structured in such a way. In an empirical study Olesen et al. (2008) recorded the day-


to-day dynamics of an arctic pollination network to analyze the attachment process of

new species (plant species that initiated flowering and pollinator species that started

to actively forage) to the network over two seasons. Studying plant–pollinator

networks during real succession, as done in the present paper, represents a new

approach to get a deeper insight into network assembly processes.

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

Understanding the patterns of specialization and generalization in plant–

pollinator interactions is a central and hotly debated topic in pollination ecology and

plant–pollinator network research (Waser et al. 1996, Johnson and Steiner 2000,

Waser and Ollerton 2006, Blüthgen et al. 2008, Petanidou et al. 2008). However,

there is a paucity of theory predicting the determinants of the degree of specialization

in pollination systems and how the latter may change during assembly processes

(Johnson and Steiner 2000). Successional status has been considered as such a

determinant (Parrish and Bazzaz 1979). From a plant’s perspective, niche theory

predicts that mature plant communities should be more specialized (have narrower

pollinator niches) than species of early successional communities, because the

expected higher density and diversity of plant species should result in more

interspecific competition and thus reduced niche overlap, at least in saturated systems

(Parrish and Bazzaz 1979). Furthermore, higher unpredictability of pollinator

availability should favor more generalized use of pollinators by plants (Waser et al.

1996) in early- as compared with late-successional plant communities.

In contrast to plants, pollinators only profit from increased specialization if

this is associated with increased resource-use efficiency or reduced interspecific

competition, e.g. through reduced handling time (e.g. Heinrich 1976, Pyke 1984).

However, if resources are similar and accessible, optimal foraging theory predicts

that, with an expected increase in resource-plant diversity during succession,


pollinators should be more generalized in mature stages to minimize foraging time

(Heinrich 1976, Pyke 1984, Waser et al. 1996, Blüthgen et al. 2007). Furthermore,

optimal foraging theory predicts that a decrease in resources provided by single plant

species should result in an increase in diet breadth (MacArthur and Pianka 1966).

Furthermore, increasing pollinator density during succession may result in a more

rapid depletion of floral resources and, as a consequence, also lead to a higher diet

breadth of individual pollinators (Fontaine et al. 2008).

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

In this paper we quantify plant–pollinator interactions at seven sites along a

chronosequence from 8 to 130 years since deglaciation at the Morteratsch glacier

foreland (SE-Switzerland). We examine how (1) plant diversity and flower

abundance, (2) pollinator diversity and visitation rate, (3) diversity of plant–pollinator

interactions, and (4) different aspects of specialization in plant–pollinator interactions

change along the chronosequence of the glacier foreland.

Material and Methods

Study site

The glacier “Morteratsch” is located at the south-facing slope of the Bernina Range in

the subalpine belt of the Central Alps of Switzerland (Fig. 1). Its foreland (9°56' E,

46°26' N) is approximately 2.2 km long and extends over altitudes from 1900 to 2010

m above sea level. Since the beginning of monitoring the glacier dynamics in 1881,

the glacier has continuously retreated at a rate of approximately 18 m per year (VAW

2007). The entire glacier foreland lies below the upper tree line of Larix decidua

Miller and Pinus cembra L. forest which reaches approximately 2400 m above sea

level in the study region. The plant primary succession along the chronosequence of

the glacier foreland is dominated by herbaceous plants in a pioneer stage, followed


after about 50 years by a stage with shrubs (Ziefle 2006). An open forest stage is

reached after around 150 years. Annual mean temperature at Samedan, 10 km north-

west of the glacier foreland at an altitude of 1700 m above sea level, is 1.2 °C, and

yearly average precipitation is 700 mm. Snow cover on the glacier foreland usually

lasts from October to May.

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

The foreland of the Morteratsch glacier was selected as model system for this

study because pronounced side moraines separate the glacier foreland from the

surrounding vegetated slopes, thus reducing plant propagule pressure from the slopes,

and because the history of the glacier retreat since 1857 is well documented (Elsener

2006 and references therein). Due to these factors the foreland of the Morteratsch

glacier has been a preferred study site for previous research on plant succession

(Ziefle 2006).

Study design

Plant–pollinator interactions were studied along 14 transects located at 7 sites of

increasing chronological age across the Morteratsch glacier foreland. Maps providing

exact information on past glacier movements and snout positions (Elsener 2006) were

digitalized and imported into a GIS. Next, the entire chronosequence from the last

glacier expansion in 1878 until present was subdivided into 7 strata limited by

isochrones of age since deglaciation, separated by time steps of roughly 20 years. The

isoclines represent the glacier extensions from 1878 until 2000 (Fig. 2). This map was

then merged with a map providing data about the characteristic vegetation for each of

110 intersection points of grid with a mesh width of 40 m and covering the entire

glacier foreland (Ziefle 2006, Elsener 2006). Then the known flooding areas close to

the glacial stream were excluded as potential areas for choosing sites. Finally, the


most representative vegetation unit was identified for each site as the unit with the

most intersection points per stratum.

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

At each site two transects that were perpendicular to each other were

established at a randomly selected intersection point of the most representative

vegetation nearby the delimiting isocline of the corresponding stratum. The seven

sites in 2008 corresponded to a chronological age since deglaciation of 8, 28, 49, 63,

84, 109 and 130 years (Fig. 2). Each intersection point was the center of a cross

consisting of a 50-m long transect in the east–west direction and a 30-m long transect

in the north–south direction. At the second oldest site (109 years since deglaciation),

sampling was not possible along the shorter transect due to very dense tree and shrub

vegetation (dominated by Betula sp., Salix sp. and Rhododendron ferrugineum L.).

Therefore, only the longer transect was sampled and analyzed at this site. Each

transect was marked with cord during the entire sampling season.

Sampling protocol for plant–pollinator interactions

Plant–pollinator interactions were observed during the entire flowering season (10

June to 21 August 2008) at roughly 2-week intervals. This resulted in seven sampling

rounds for each transect. The snow melted at roughly the same time at each sampling

site. Flower-visiting insects were sampled between 9:30 and 17:30 h in sunny weather

conditions. Because of potential temporal differences in pollinator activity, each site

was sampled at least twice in the morning (9:30–12:30), twice in the early afternoon

(12.30–15:30) and twice in the late afternoon (15:30–17:30). Time of day in which

sampling took place was varied randomly across sites and sampling rounds. During

each sampling round, all flowering vascular plants and their insect visitors were

sampled within a belt of 1 m along the transects. While slowly walking along the


transects every insect visitor landing on an open flower was caught with a sweep-net

immediately after landing and put into a 50 ml tube containing ethyl acetate. Thus,

the number of individual flowers within the same inflorescence that an insect visited

was not considered in the analyses. Each transect was walked twice during each

sampling round. The time spent for each transect and sampling round was 15 min for

the short transects and 25 min for the long transect, resulting in total sampling time of

4 hours and 40 min for each site. If possible, flower visitors were identified to species

level, otherwise to genus or family level (see Appendix 1). Visitation rate was defined

as the total number of flower visits observed along a transect during one sampling

round. Flower visitors were assumed to be pollinators. Flower abundance of vascular

plants was recorded as the number of inflorescences of each flowering plant species,

including species that did not receive any pollinator visits.

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

Quantifying plant–pollinator network structure

The pooled, quantified plant–pollinator networks for each site were plotted. To detect

changes in the interaction structure of the plant–pollinator networks along the

chronosequence, the following quantitative, weighted properties (i.e. properties that

account for variation in interaction strength) were calculated for each transect

(sampling rounds pooled): linkage density (LDq), interaction diversity (IDq),

pollinator generality (Gpoll,q, the mean number of plant species visited by a pollinator

species weighted by interaction strength; also referred to as generalization in the

following text) and plant generality (Gplant,q, the mean number of pollinator species

visiting a plant species weighted by interaction strength, equivalent to vulnerability

Vq of Bersier et al. 2002). Plant species that were not visited by pollinators were not

included in these calculations. These metrics were calculated following Bersier et al.


221

222

223

(2002) and Albrecht et al. (2007). They are based upon the Shannon indices of

entropy to measure the diversity of flowering plants (HN) and pollinator individuals

(HP) for each taxon k:

Error! Bookmark not defined.HN,k = −

bik

b• ki=1

s

∑ log2bik

b• k 224

HP,k = −bkj

bk•j=1

s

∑ log2

bkj

bk• 225

226

227

228

229

230

231

Here b•k is the number of flowering plants visited by visitor species k, and bk• is the

number of flower visitors visiting plant species k. The network metrics are based on

the “reciprocals” of HN,k and HP,k (Bersier et al. 2002):

nN,k =2HN,k

0

⎧⎨⎪

⎩⎪if b•k = 0

nP,k =2HP,k

0

⎧⎨⎪

⎩⎪if bk• = 0

The weighted linkage density (LDq) was calculated for each) as:

LDq =12

bk•

b••

nP,k+b• k

b••

nN,kk=1

s

∑k=1

s

∑⎛⎝⎜

⎞⎠⎟ 232

233

234

where b•• is the total number of visited flowering plants. Weighted pollinator

generality (Gpoll,q) and plant generality (Gplant,q) were calculated as:

Gpoll,q =b• k

b••k=1

s

∑ nN,k

235

Gpoll,q =bk•

b••k=1

s

∑ nP,k

236

237

238

239

To measure network level generalization, we calculated the recently

developed index H2’ by Blüthgen et al. (2006). This index characterizes the degree of

specialization among the two trophic levels in bipartite networks as the deviation


240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

from an expected probability distribution of interactions, evaluated by the

standardized two-dimensional entropy. For a detailed description of the mathematical

derivation of the index see Blüthgen et al. (2006). The index accounts for variability

in interaction frequencies and is not only robust against variation in the size of a

network, but also its shape, i.e. the skewness of the frequency distribution of the

recorded species (Blüthgen et al. 2006, 2008).

An additional aspect of generalization in weighted, bipartite networks is

interaction strength asymmetry, calculated as the grand mean of the imbalance

between the interaction strengths of each species pair within a network (Bascompte et

al. 2006). We used the modified version of interaction strength asymmetry proposed

by Dormann et al. (2009), in which all singleton species are omitted from the network

to avoid that these species receive a very high influence.

The measure of the diversity of interactions (IDq) was calculated using the

Shannon index:

IDq = pi∑ log2 pi( )

where pi is the proportion of the interaction i of the total number of interactions.

Furthermore, we adapted Hurlbert’s probability of inter-specific encounters (PIE) as a

measure for evenness in ecological communities, which is unbiased by the number of

species in a sample (Hurlbert 1971), to measure interaction evenness across networks

as:

PIE =L

L+1⎡⎣⎢

⎤⎦⎥

1−Li

L⎛⎝⎜

⎞⎠⎟

2

i∑

⎡

⎣⎢⎢

⎤

⎦⎥⎥ 260

261

262

263

where L is the total number of individual plant–pollinator interactions and Li the

number of individual interactions of the i-th link between a specific plant–pollinator

pair.


We believe that the above mentioned weighted network properties are the

most appropriate metrics currently available to investigate our research questions

(Bersier et al. 2002, Blüthgen 2006, Dormann et al. 2009). However, in order to make

it easier to compare the network properties of this study with those of other plant–

pollinator network studies that give only qualitative network metrics we also

calculated these traditionally used, unweighted properties (number of links per

species, connectance, average pollinator species degree, average plant species degree,

average network level degree and nestedness). Connectance is the realized proportion

of all possible links. Species degree is the number of different species a species

interacts with. Nestedness is a measure of departure from systematic arrangement of

species by niche width (Bascompte et al. 2003). The nestedness “temperature” T (0°–

100°) measures the departure from a perfectly nested interaction matrix, with

maximum nestedness defined as T = 0°.

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

For the calculation of the more complex network properties H2’, interaction

strength asymmetry and nestedness the data of the two transects and the seven

sampling rounds were pooled for each of the seven sites (to avoid small network sizes

potentially affecting the functioning of the more complex algorithms used for the

calculations of these metrics). To test whether pooling of the two transects of a site

would lead to different results for other measures, all analyses for those other

measures were also performed using these pooled data (fitting linear models with the

single continuous explanatory variable site age). The results of these analyses were

very similar to those obtained by the mixed-effect model analyses presented here. H2’

was calculated on the webpage http://nils.mib.man.ac.uk/~nils/stat/. For the

calculations of all other network metrics (except Hurlbert’s PIE) the bipartite package

supplied in the R-system of statistical computing was used (Dormann et al. 2009).

286

287

288

http://nils.mib.man.ac.uk/%7Enils/stat/


289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

Data analysis

We analyzed variation in the dependent variables of plant and pollinator species

richness, plant and pollinator Shannon diversity indices, flower abundance, visitation

rate of pollinators and the above-mentioned network properties with linear mixed-

effects models. For the Shannon diversity indices we used the formula described

above for interaction diversity IDq, but with pi as the proportion of inflorescences or

flower visiting individuals belonging to the i-th plant or pollinator species,

respectively. Site age (number of years since deglaciation) was treated as fixed

continuous explanatory variable. The remaining variation among sites (i.e. site fitted

after site age), transect nested within site, sampling round and the site x sampling

round interaction were treated as random effects. For the analyses of the network

properties data from all sampling rounds were pooled, because some networks of

single sampling rounds were too small for a reliable analysis (thus, sampling round

was not included in the analyses of network properties). According to the

specification of the random terms for the mixed-model analysis the fixed term site age

was tested against the residual variation among sites. This avoided a potential pseudo-

replication problem that would have occurred if site age would have been tested

against variation among transects or among residuals (= transect x sampling round

interactions). Transect length was included as a covariate in the models to account for

any potential effects of transect length. Finally, to test whether variation in pollinator

species richness and visitation rate along the chronosequence were explained by

variation in plant species richness and flower abundance, these two measures were

used as further covariates in the models with pollinator visitation rate, pollinator


species richness and pollinator diversity as the dependent variables and sampling

round, site age and transect length as explanatory variables.

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

We analyzed similarity between the networks at the different sites in observed

plant–pollinator interactions by calculating the modified Simpson’s index proposed

by Koleff et al. (2003). This index measures the similarity once differences in number

of interactions have been removed. It is calculated as the number of interactions of

two sites divided by the minimum number of interactions present in one of the two

sites (Petanidou et al. 2008). This modified Simpson’s index for interaction similarity

was calculated for each of the 7 x 6 / 2 = 21 site pairs. To test whether plant–

pollinator networks of closer sites were more similar in interactions than those of sites

further apart from each other along the chronosequence we fitted a linear model with

the response variable modified Simpson’s index for interactions and the explanatory

variables closeness (in years since deglaciation, i.e. 20 years for the pair consisting of

the 8- and 28-year old sites), mean age of the site pair (i.e. 18 years for the

aforementioned site pair), indicating whether similarity changed along the

chronosequence, and their interaction.

Mixed effect models were fitted using the lme-function of the nlme package

supplied in the R-system of statistical computing (R Development Core Team 2008).

Arithmetic means ± 1 standard error (SE) are reported. Due to the low number of sites

the power of the F1,5-test for site age was low and thus marginal significances at P <

0.1 are also reported as a protection against making too many type-II errors (Toft and

Shea 1983). The effect sizes in these analyses can easily be calculated from the

reported F-values and the degrees of freedom by taking the square-root of the ratio r2

= (df numerator * F) / (df numerator * F + df denominator) (Rosenthal and Rosnow


1985), thus the square-root of F / (F + 5). For example, an F-value of 5 results in an

effect size of 0.50, whereas an F-value of 10 results in an effects size of 0.67.

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

Results

Pooled, quantified plant–pollinator networks for all sites along the chronosequence

are shown in Figure 2. Over the entire season a total of 54 plant species were

flowering and 37 of them (67%) were visited by a total of 92 pollinator species (Fig.

2; Appendix 1). In total, 506 plant–pollinator interactions (pollinator visits) were

recorded. Whereas flower abundance was not significantly higher at late- than at

early-succesional sites (F1,5 = 3.98, P = 0.103; Fig. 3a), species richness of flowering

plants (F1,5 = 12.89, P = 0.016), Shannon diversity of flowering plants (F1,5 = 8.77, P

= 0.032), visitation rate (F1,5 = 8.47, P = 0.033), species richness of pollinators (F1,5 =

8.37, P = 0.034) and network size (total number of interacting plant and pollinator

species; F1,5 = 7.87, P = 0.038) all increased along the chronosequence (Fig. 3b–e).

Species richness and Shannon diversity of pollinators peaked at intermediate

successional age (84 years since deglaciation; Fig. 3e,f).

To test if the observed pattern in pollinator species richness and visitation rate

along the chronosquence was related to plant species richness and flower abundance,

we introduced these covariates stepwise into the linear models. Because plant species

richness and flower abundance were highly correlated (F1,6 = 58.84, P < 0.001), we

only report results of the stronger explanatory variable, which was plant species

richness. Plant species richness was highly significant in explaining variation in

pollinator richness (F1,76 = 52.78, P < 0.001) and pollinator visitation rate (F1,76 =

48.83, P < 0.001). Moreover, site age was not significant anymore in explaining

variation in pollinator richness or visitation rate when plant species richness was


included in the model (results not shown). This suggests that increasing site age

influenced pollinators indirectly via its effects on plant species richness, although

pollinator species richness and pollinator diversity showed a saturating or peaked

pattern, respectively, while pollinator visitation rate continued to increase in the late-

succesional stages.

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

Diptera accounted for 74%, Hymenoptera (almost exclusively Apidae)

accounted for 24% and other insect orders accounted for 2% of the visits to flowers

(Fig. 2, Appendix 2). The number of visits by Diptera significantly increased along

the chronosequence (F1,5 = 14.77, P = 0.012), whereas that of Hymenoptera showed a

non-significant decrease (F1,5 = 3.74, P = 0.111). This resulted in a significant change

in the relative number of visits by Diptera and Hymenoptera along the

chronosequence (decline in log-transformed Hymenoptera/Diptera-ratio: F1,5 = 14.43

P = 0.013) from the youngest site (55% Hymenoptera, 45% Diptera) to the oldest site

(4% Hymenoptera, 96% Diptera).

At the youngest two sites (8 and 28 years since deglaciation) two plant

species, Epilobium fleischeri (Onagraceae; species code 1268: Fig. 2, Appendix 2)

and Saxifraga bryoides (Saxifragaceae; species code 909: Fig. 2, Appendix 2),

received by far the most pollinator visits. Bumblebees almost exclusively visited E.

fleischeri at these two sites (Fig. 2). The interaction structure at these sites was

characterized by the dominant interaction between the bumblebee Bombus sichelii

and E. fleischeri (Fig. 2). At site 3 (49 years since deglaciation), 44% of the pollinator

individuals were recorded on Saxifraga paniculata (Saxifragaceae; species code 887:

Fig. 2, Appendix 2), which produced flowers most abundantly (34%) at this site. The

percentage of flowering plant species for which pollinator visits were observed was

lower at the two pioneer sites (24% at the 8-years old site and 31% for the 28-years


old site), while at older sites this percentage was much higher (54% on average,

ranging from 47% to 59%; Fig. 2).

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

Interaction diversity increased with site age (F1,5 = 7.38, P = 0.041), at least up

to 84 years (Fig. 4a). Interaction evenness tended to increase throughout the entire

130-year period covered by the chronosequence (F1,5 = 6.38, P = 0.053; Fig. 4b).

However, weighted linkage density (F1,5 = 2.12, P = 0.205; Fig. 4c) remained more or

less constant along the chronosequence, while the unweighted number of links per

species increased with site age (F1,5 = 8.42, P = 0.034; Appendix 2). Connectance

tended to decrease with site age (F1,5 = 4.13, P = 0.098; Appendix 2).

Weighted network-level specialization measured as Blüthgen’s H2’

significantly decreased along the chronosequence (F1,5 = 8.62, P = 0.032; Fig. 4d).

The strongest decline occurred between the 28- and the 63-year old sites. This

decrease in network-level specialization reflected the higher average number of

pollinator species visiting a plant species (increase in pollinator generality with site

age; F1,5 = 39.08, P = 0.002; Fig. 4e), while the weighted average number of plants

visited by a pollinator did not change (F1,5 = 0.12, P = 0.747; Fig. 4f). The qualitative,

unweighted equivalents of the above mentioned weighted metrics showed

qualitatively identical patterns (increase in average network level and pollinator

species degree: F1,5 = 8.04, P = 0.037; F1,5 = 7.86, P = 0.001; no directional trend in

average plant species degree: F1,5 = 1.18, P = 0.327; Appendix 2). Examples of

pollinator species that occur at most sites and that visited more plant species at older

sites compared to earlier ones include Tricops sp., Paonia sp. or Bombus sichelii. For

most of the abundantly flowering plant species the highest degree values (number of

visitor taxa per plant species) were recorded at the site with highest flower abundance

of the corresponding plant species. For example, for Epilobium fleischeri this was the


8-year old site, for Saxifraga paniculata the 49-year old site or for Achillea erba-rotta

the 84-year old site.

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

The contrasting patterns of plant and pollinator generalization in early-

successional stages resulted in a decline in interaction strength asymmetry along the

chronosequence (F1,5 = 9.39, P = 0.028; Fig. 4g), with pollinators being more

specialized in these early assembly stages than plants. Conversely, networks at

mature successional stages showed a more nested interaction structure than those at

early stages (F1,5 = 9.39, P = 0.028; Fig. 4h).

Interaction similarity between networks (the modified Simpson’s index) was

higher for closer sites than for sites further apart from each other along the

chronosequence (F1,17 = 13.31,12 P = 0.002; Appendix 3). However, average

interaction similarity did not change with site age (F1,17 = 0.04, P = 0.854; Appendix

3).

Discussion

Our study shows pronounced changes in the structure of plant–pollinator networks

along the chronosequence of a primary glacier foreland succession. These changes

can be interpreted as community assembly processes leading to increasingly diverse

plant–pollinator systems at least during the first 80 years after deglaciation. With

increasing site age, the average pollinator visited more plant species (increasing

pollinator generalization, but not plant generalization), suggesting that plant

community assembly may have been the driver in this diversity-begets-diversity

assembly process. As a consequence, network level specialization (measured as

Blüthgen’s H2’) and interaction strength asymmetry declined up to a mid-successional

stage.


437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

Plant and pollinator community assembly

Species richness and Shannon diversity of plants with open flowers increased from

the youngest to the mid- and late-successional sites along the chronosequence (see

Fig. 3). This confirms findings of previous studies on plant primary succession on the

foreland of the Morteratsch glacier (Ziefle 2006) and most other studied glacier

forelands (Matthews 1992, Walker and del Moral 2003).

Pollinator community assembly seemed to follow plant community assembly.

Thus, the increase in plant diversity along the chronosequence was paralleled by an

increase in visitation rate and species richness of pollinators and, at least up to a mid-

successional stages, pollinator Shannon diversity (see Fig. 3). Recent experimental

work also indicates that pollinator visitation rate increases linearly with plant species

richness, while pollinator diversity scales with plant species richness along a

saturating curve (Ebeling et al. 2008). Considering the relatively high mobility of

most pollinator taxa, colonization of freshly available, early-successional habitats by

pollinators was probably not dispersal limited. In addition, the increasing

generalization of pollinators suggests that the diversity of available floral resources

was the crucial factor affecting pollinator community assembly on the local scale of

our seven sites (Potts et al. 2003b). Evidence from studies on secondary successions

also indicates that pollinator diversity mainly mirrors plant diversity, which can lead

to increasing pollinator diversity with increasing age of meadows (Gathmann et al.

1994) or decreasing pollinator diversity along recovering vegetation gradients after

fire, with highest plant diversity levels in the first few years after a burn (Potts et al.

2003a).


In addition to the changes in pollinator species richness, the species

composition of pollinator communities also changed along the chronosequence of the

primary succession of our glacier foreland. Bees, particularly bumblebees, dominated

in pioneer stages and flies in mature stages. These changes in pollinator species

composition probably reflect those in species composition of the plant communities

along the chronosequence. For example, the abundance of bumblebees closely

mirrors the flower abundance of Epilobium fleischeri, a pioneer species with highest

abundances at the youngest two sites.

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

Plant–pollinator network assembly

Interaction diversity and less so interaction evenness increased along the

chronosequence up to mid-successional stages. These trends were consistent with an

increase in pollinator generality and a decrease in network-level specialization along

the chronosequence (see Fig. 4). In contrast to some measures used previously to

characterize generalization or specialization, such as the average species degree

(number of taxa interacting with one partner species; e.g. Herrera 2005) or

connectance (proportion of realized of all possible links; Jordano 1987, Olesen and

Jordano 2002), the quantitative, weighted network metrics used in this study are

largely independent of network size and sampling effort (Bersier et al. 2002, Blüthgen

et al. 2006). The metric H2’ additionally controls for variation in skewness of the

frequency distribution of the data, which may affect some weighted properties such as

generality or interaction strength asymmetry (Blüthgen et al. 2008). In spite of their

deficiencies, the qualitative metrics of plant and pollinator generality and network

level generalization (average plant, pollinator and network level degree) produced

similar results as their quantitative, weigthed counterparts. Only for connectance, a


qualitative metric that has often been used to describe generalization in networks,

there was a tendency to decline with site age, while the results for quantitative H2’

(and the qualitative average network degree) indicated a significant increase in

network level generalization. Connectance as a qualitative metric to characterize

generalization across networks has been criticized (Blüthgen et al. 2006, 2008 and

references therein) because it is not only sensitive to sampling intensity but also

because it is negatively correlated with network size (e.g. Olesen and Jordano 2002).

Since network size increased with site age in our study, a decrease in connectance

with site age is expected only due to this mathematically inherent scale dependence

(Blüthgen et al. 2006). This emphasizes the importance of using properties that are

more robust against such effects when comparing networks differing in size (Bersier

et al. 2002, Blüthgen et al. 2006).

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

Why does plant generality not decrease with successional age?

Our results do not support niche-theoretical predictions that plant species in mature

communities should be more specialized with respect to their pollinators than those in

early-successional communities because of increased interspecific competition and

reduced niche overlap (Parrish and Bazzaz 1979). Although some characteristic plant

species of the two pioneer sites, such as Epilobium fleischeri or Saxifraga bryoides

were visited by a relatively high number of different pollinator species (Fig. 2),

overall plant generalization did not show a significant directional trend along the

chronosequence. Interestingly, however, some abundantly flowering plant species

such as Arabis alpina at the pioneer sites received no or very few pollinator visits.

Most studies on plant–pollinator networks do not provide data regarding the

percentage of flowering plant species for which no pollinator visits are observed. In


our study, approximately one third of all flowering species received no pollinator

visits, although most of them are known to be normally pollinated by insects. This

percentage was roughly twice as high for the two youngest sites compared with the

more mature ones, indicating that a considerably lower number of flowering plant

species is integrated into the pollination network at early than at late stages of

network assembly.

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

Why does pollinator generality increase with successional age?

Because plant diversity increased yet total flower abundance remained relatively

constant along the chronosequence, the abundance of flowers of individual plant

species declined. Under these circumstances and according to optimal foraging

theory, pollinators should become generalists to minimize foraging time (MacArthur

and Pianka 1966, Heinrich 1976, Pyke 1984). Thus, the increase in pollinator

generalization and the decrease in network-level specialization found in this study

may reflect the increase in heterospecific floral resources during the primary

succession. Another prediction of optimal foraging theory that is supported by our

findings is that the increase in pollinator abundance during the succession, indicated

by the positive correlation between visitation rate and site age, should result in a

higher diet breadth of individual pollinators as a response to more rapid depletion of

floral resources at high pollinator abundance (Fontaine et al. 2008). Such a parallel

increase in abundance and generalization along the chronosequence was found for

example for several taxa of flies or the bumblebee species Bombus sichelii. However,

also many plant species showed a positive correlation between flower abundance and

generalization level (see Fig. 4). A simple mechanism of network assembly that is

compatible with such a pattern is the preferential attachment model (Barabasi and


Albert 1999). According to this model, new species have a higher probability to link

to species that have already many links in a network, leading to a “rich-get-richer”

process. Studying the day-to-day attachment of new species (i.e. species that started

to flower or actively forage) to an arctic plant–pollinator network during two seasons,

Olesen et al. (2008) found that new species indeed tended to attach to already well-

connected species, although this was constrained by some morphological and

phenological mismatching of plants and pollinators. Although we did not study such

factors as potential drivers of the observed patterns of network assembly in the

present study, we found at least some evidence that abundance was positively linked

with the degree of many plant and pollinator species.

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

Our findings are consistent with the view that plant–pollinator networks, at

least in later successional stages, may be not so much niche-structured, but may rather

reflect a high degree of opportunism with respect to resource use and thus a more

neutral way of plant–pollinator networks assembly (Vazquez 2005, Petanidou et al.

2008). Such opportunism is, however, obviously limited by phenological,

morphological and co-evolutionary constraints (Jordano et al. 2003, Stang et al. 2006,

Bascompte and Jordano 2007, Santamaría and Rodríguez-Gironés 2007, Vazquez et

al. 2009a,b). On glacier forelands and other systems structured by primary

succession, plants often occur in scattered and sparse populations, in particular during

pioneer stages, and thus may depend much more than those of later assembly stages

on specialized pollinators that show high levels of fidelity (Johnson and Steiner 2000

and references therein). On the other hand, a plant species typically found in early

stages of primary succession that relies to a large degree on animal pollination, such

as E. fleischeri (Albrecht, unpublished data), should attract a diverse pollinator

community considering the unpredictability of pollinator availability during this stage


(Parrish and Bazzaz 1979). This should lead to exactly the pattern found here, with

higher pollinator specialization but not plant specialization in the pioneer stages of the

succession, resulting in a higher interaction strength asymmetry in these early stages.

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

Increased nestedness of plant–pollinator networks with successional age

Most plant–pollinator networks that have been studied so far were characterized by a

highly nested interaction structure, i.e. more specialized species tend to interact with

interaction partners that are highly generalized, and with nestedness levels

significantly higher compared with those of null models (Bascompte et al. 2003,

Bascompte and Jordano 2007). Our finding of high nestedness levels in the mature

assembly stages is in a line with the existing studies typically investigating mature,

relatively stable systems. The early assembly stages, however, were significantly less

nested, indicating that nestedness is a typical feature of rather developed and stable

plant–pollinator networks, but not necessarily of still evolving ones at the beginning

of assembly processes. This view would also be in agreement with simulation models

suggesting that a highly nested interaction structure makes ecological networks more

cohesive and stable (Bascompte et al. 2003, 2009, Bascompte and Jordano 2007).

Conclusions

It is a well-known fact that ecological network data so far have always been

constructed of incomplete samples (Blüthgen et al. 2008). Also in the present study,

sampling thoroughness may not have been extremely high. However, here we were

primarily interested in the relative change of network properties and their pattern

along a primary successional gradient rather than estimate values, for example of

fundamental specialization, most accurately. Despite potential limitations inherent in


the space-for-time substitution approach and the above mentioned uncertainties in

sampling thoroughness typical for ecological network data, the analysis of networks

along chronosequences provide unique insights into the assembly of multitrophic

communities and their interaction structure that cannot be gained by short-term

experimental studies (Matthews 1992, Hodkinson et al. 2004). To our knowledge, this

is the first study that used a network approach to explore plant–pollinator assembly

during primary succession. We found pronounced changes in the interaction structure

of the plant–pollinator communities during the studied glacier foreland succession,

which was characterized by a strong increase in pollinator generality, but not plant

generality, during network assembly, resulting in a decline of network level

specialization and interaction strength asymmetry. An important implication of this

study is that the network structure of evolving ecological communities in early

assembly stages can be significantly different to that of the usually studied, rather

mature and stable ones. Further research is needed to unravel the exact mechanistic

processes underlying the observed patterns in the evolution of plant–pollinator

network properties during this type of primary succession. Ecological network

analysis along successional gradients is a promising tool to gain new insights in how

networks of multitrophic communities assemble.

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

Acknowledgements

We dedicate this paper to Christine Müller, who died after the start of this study; she

was a great inspiration for us. We thank Ulrich Schneppat for valuable advice on

insect preparation and the following taxonomic specialists for help with the species

identification: Jean-Paul Haenni (Diptera) and Hermann Blöchlinger (Diptera), Ruth

Bärfuss (Syrphidae), Rainer Neumeyer and Felix Amiet (Apidae), Seraina Klopfstein


and Hannes Baur (Ichnneumonidae), Jürg Schmid (Lepidoptera), Peter Herger

(Coleoptera) and Denise Wyniger (Heteroptera). We are grateful to Stefan Elsener for

providing GIS data and Roland Schmid for help with drawing the plant–pollinator

networks. We also thank Martina Stang and two anonymous reviewers for their

valuable comments on an earlier version of this manuscript. This work was supported

by the Swiss National Science Foundation (grant 631-065950 to Christine Müller).

611

612

613

614

615

616


References 617

618

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

Albrecht, M. et al. 2007. Interaction diversity within quantified insect food webs in

restored and adjacent intensively managed meadows. – J. Anim. Ecol. 76: 1015–

1025.

Bascompte, J. et al. 2003. The nested assembly of plant–animal mutualistic networks.

– Proc. Natl Acad. Sci. 100: 9383–9387.

Bascompte, J. et al. 2006. Asymmetric coevolutionary networks facilitate biodiversity

maintenance. – Science 312: 431–433.

Bascompte, J. and Jordano, P. 2007. Plant–animal mutualistic networks: the

architecture of biodiversity. – Annu. Rev. Ecol. Evol. Syst. 38: 567–593.

Bascompte J. and Stouffer, D. B. 2009. The assembly and disassembly of ecological

networks. – Phil. Trans. R. Soc. B 364: 1781–1787.

Bersier, L. F. et al. 2002. Quantitative descriptors of food-web matrices. – Ecology

83: 2394–2407.

Blüthgen, N. et al. 2006. Measuring specialization in species interaction networks. –

BMC Ecol. 6: 9.

Blüthgen, N. et al. 2007. Specialization, constraints, and conflicting interests in

mutualistic networks. – Curr. Biol. 17: 341–346.

Blüthgen, N. et al. 2008. What do interaction network metrics tell us about

specialization and biological traits? – Ecology 89: 3387–3399.

Clements, F. E. 1916. Plant succession. Carnegie Institute of Washington.

Dormann, C. et al. 2009. Indices, graphs and null models: analyzing bipartite

ecological networks. – The Open Ecology Journal 2: 7-24.

Ebeling, A. et al. 2008. How does plant richness affect pollinator richness and

temporal stability of flower visits? – Oikos 117: 1808–1815.


Elsener, S. 2006. Besiedlungsdynamik von Arve (Pinus cembra L.) und Lärche (Larix

decidua Mill.) im Vorfeld des Morteratschgletschers. – Diploma thesis. Swiss

Federal Institute for Forest, Snow and Landscape Research.

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

Fontaine, C. et al. 2008. Generalist foraging of pollinators: diet expansion at high

density. – J. Ecol. 96: 1002–1010.

Foster, B. L. and Tilman, D. 2000. Dynamic and static views of succession: testing

the descriptive power of the chronosequence approach. – Plant Ecol. 146: 1–10.

Gathmann, A. et al. 1994. Trap-nesting bees and wasps colonizing set-aside fields –

succession and body-size, management by cutting and sowing. – Oecologia 98: 8–

14.

Heinrich, B. 1976. The foraging specializations of individual bumblebees. – Ecol.

Monogr. 146: 105–128.

Herrera, C. M. 2005. Plant generalization on pollinators: Species property or local

phenomenon? – Am. J. Bot. 92: 13–20.

Hodkinson, I. D. et al. 2001. What a wonderful web they weave: spiders, nutrient

capture and early ecosystem development in the high Arctic – some counter-

intuitive ideas on community assembly. – Oikos 95: 349–352.

Hodkinson, I. D. et al. 2004. Invertebrate community assembly along proglacial

chronosequences in the high Arctic. – J. Anim. Ecol. 73: 556–568.

Hubell, S. P. 2001. The unified neutral theory of biodiversity and biogeography. –

Princeton Univ. Press.

Hurlbert, S. 1971. Nonconcept of species diversity – critique and alternative

parameters. – Ecology 52: 577–586.

Ings, T. C. et al. 2009. Ecological networks – beyond food webs. – J. Anim. Ecol. 78:

253–269.


Johnson, S. D. and Steiner, K. E. 2000. Generalization versus specialization in plant

pollination systems. – Trends Ecol. Evol. 15: 140–143.

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

685

686

687

688

689

Jordano, P. 1987. Patterns of mutualistic interactions in pollination and seed dispersal:

connectance, dependence asymmetries, and coevolution. – Am. Nat. 129: 657.

Jordano, P. et al. 2003. Invariant properties in coevolutionary networks of plant–

animal interactions. – Ecol. Lett. 6: 69–81.

Kaufmann, R. 2001. Invertebrate succession on an alpine glacier foreland. – Ecology

82: 2261–2278.

Koleff, P. 2003. Measuring beta diversity for presence-absence data. – J. Anim. Ecol.

72: 367-382.

Lortie, C. J. et al. 2004. Rethinking plant community theory. – Oikos 107: 433–438.

MacArthur, R. H. and Pianka, E. R. 1966. On optimal use of a patchy environment. –

Am. Nat. 100: 603-609.

Matthews, J. A. 1992. The ecology of recently deglaciated terrain. – Cambridge Univ.

Press.

Olesen, J. M. and Jordano, P. 2002. Geographic patterns in plant–pollinator

mutualistic networks. – Ecology 83: 2416–2424.

Olesen, J. M. et al. 2007. The modularity of pollination networks. – Proc. Natl Acad.

Sci. 104: 19891–19896.

Olesen, J. M. et al. 2008. Temporal dynamics in a pollination network. – Ecology 89:

1573–1582.

Parrish, J. A. D. and Bazzaz, F. A. 1979. Difference in pollination niche relationships

in early and late successional plant communities. – Ecology 60: 597–610.


Petanidou, T. et al. 2008. Long-term observation of a pollination network: fluctuation

in species and interactions, relative invariance of network structure and

implications for estimates of specialization. – Ecol. Lett. 11: 564–575.

690

691

692

693

694

695

696

697

698

699

700

701

702

703

704

705

706

707

708

709

710

711

712

713

714

Potts, S. G. et al. 2003a. Response of plant–pollinator communities to fire: changes in

diversity, abundance and floral reward structure. – Oikos 101: 103–112.

Potts, S. G. et al. 2003b. Linking bees and flowers: how do floral communities

structure pollinator communities? – Ecology 84: 2628–2642.

Proulx, S. R. et al. 2005. Network thinking in ecology and evolution. – Trends Ecol.

Evol. 20: 345–353.

Pyke, G. H. 1984. Optimal foraging theory – a critical review. – Annu. Rev. Ecol.

Syst. 15: 523–575.

Rosenthal, R. and Rosnow, R. L. 1985. Contrast analysis: Focused comparisons in the

analysis of variance. – Cambridge Univ. Press.

Santamaría, L. and Rodríguez-Gironés, M. 2007. Linkage rules for plant–pollinator

networks: Trait complementarity or exploitation barriers? – PLoS Biol. 5: e31.

Sargent, R. D. and Ackerly, D. D. 2008. Plant–pollinator interactions and the

assembly of plant communities. – Trends Ecol. Evol. 23: 123–130.

Stang, M. et al. 2006. Size constraints and flower abundance determine the number of

interactions in a plant–flower visitor web. – Oikos 112: 111–121.

Toft, C. A. and Shea, P. J. 1983. Detecting community-wide patterns: estimating

power strengthens statistical inference. – Am. Nat. 122: 618-625.

VAW 2007. Gletscherberichte (1881–2007). – Laboratory of Hydraulics, Hydrology

and Glaciology (VAW).

Vazquez, D. P. and Aizen, M. A. 2003. Null model analyses of specialization in

plant–pollinator interactions. – Ecology 84: 2493–2501.


Vazquez, D. P. and Aizen, M. A. 2004. Asymmetric specialization: a pervasive

feature of plant–pollinator interactions. – Ecology 85: 1251–1257.

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

Vazquez, D. P. 2005. Degree distribution in plant–animal mutualistic networks:

forbidden links or random interactions? – Oikos 108: 421–426.

Vazquez, D. P. et al. 2009a. Uniting pattern and process in plant–animal mutualistic

networks: a review. – Ann. Bot. 103: 1445–1457.

Vazquez, D. P. et al. 2009b. Evaluating multiple determinants of the structure of

plant–animal mutualistic networks. – Ecology 90: 2039–2046.

Walker, L. R. and del Moral, R. 2003. Primary succession and ecosystem

rehabilitation. – Cambridge Univ. Press.

Waser, N. M. et al. 1996. Generalization in pollination systems, and why it matters. –

Ecology 77: 1043–1060.

Waser, N. M. and Ollerton, J. 2006. Plant–pollinator Interactions. – Univ. of Chicago

Press.

Ziefle, M. 2006. Gletscherschwund und Vegetationsdynamik in Morteratsch

(Pontresina, Schweiz). – Diploma thesis. Univ. of Zürich.


Figure legends 733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

Fig. 1. Location and map of the Morteratsch glacier foreland at the south-facing slope

of the Bernina Range (Engadine, Switzerland). Positions of sampling sites

representing a chronosequence from 8 to 130 years since deglaciation together with

isoclines of glacier extensions from 1857 to 2008 (grey lines) and the glacial stream

(black line).

Fig. 2. Quantified plant–pollinator networks at seven sites of increasing distance from

the glacier snout across the Morteratsch glacier foreland, representing a

chronosequence from 8 to 130 years since deglaciation. Lower bars represent flower

abundance of vascular plants and upper bars pollinator visitation rates drawn at

different scales. The width at the basis of the wedges represents interaction frequency.

Taxa are coded by numbers (see Appendix 2 for names).

Fig. 3. Flower abundance (a), plant species richness (b), plant diversity (Shannon

diversity index) (c), pollinator visitation rate (d), pollinator species richness (e) and

pollinator diversity (Shannon diversity index) (f) at seven sites of increasing distance

from the glacier snout across the Morteratsch glacier foreland, representing a

chronosequence from 8 to 130 years since deglaciation. The broken lines connect the

mean values of the two transects of each site (for better visualisation they are slightly

jittered; at the 109-year old site only one transect could be sampled). For the

calculation of (a) to (c), all flowering plant species, including species that did not

receive any pollinator visits, were used. Plant–pollinator interactions were sampled

during a total of four hours and 40 min per site during seven sampling rounds from 10

June to 21 August 2008.


758

759

760

761

762

763

764

765

766

767

768

769

770

771

Fig. 4. Changes in quantitative, weighted network properties at seven sites of

increasing distance from the glacier snout across the Morteratsch glacier foreland,

representing a chronosequence from 8 to 130 years since deglaciation: (a) interaction

diversity (Shannon diversity index), (b) interaction evenness (measured as Hurlbert’s

probability of interspecific encounters), (c) linkage density, (d) network level

specialization (measured as Blüthgen’s H2’ using the pooled networks of a site, see

“Material and Methods”), (e) pollinator generality and (f) plant generality. The broken

lines connect the mean values of the two transects of each site (for better visualisation

they are slightly jittered; at the 109-year old site only one transect could be sampled).

Plant species that were not visited by pollinators were not included in these analyses.

For description and calculation of the network properties see “Material and Methods”

section.


Figures 772

773

774

Figure 1

775


Figure 2 776

777

778

779


Figure 3 780

781

782

783

784

785

786

787

788

789

790

791

792

793

794

795

796

797


Figure 4 798

799

800

801


Appendix 1. List of codes for pollinator and plant species used in Fig. 2. 802

Code Pollinator Code Plant 3001 Anthidium montanum Morawitz 1864 221 Papaver croceum Ledebour 3002 Anthomyiidae sp. 385 Cerastium fontanum ssp. fontanum Baumgarten

3003 Bombus monticola Smith 1849 390 Cerastium arvense ssp. strictum Koch

3004 Bombus mucidus Gerstaecker 1869 419 Silene vulgaris ssp. vulgaris Garcke 3005 Bombus pyrenaeus Perez 1879 447 Oxyria digyna Hill

3006 Bombus sichelii Radoszkowski 1859 472 Rumex scutatus L.

3007 Bombus sp. 583 Salix sp. 3008 Calliphoridae sp. 652 Cardamine resedifolia L.

3009 Cantharis pagana Rosenhauer 1847 667 Arabis alpina L.

3010 Cauchas fibulella Denis & Schiffermüller 1775 778 Myrica germanica L. 3011 Chamaesyrphus scaevoides Fallén 1817 786 Rhododendron ferrugineum L.

3012 Cheilosia albitarsis Meigen 1822 789 Vaccinium vitis–idaea L.

3013 Cheilosia sp. Meigen 1822 803 Pyrola minor L. 3014 Cheilosia variabilis Panzer 1798 818 Primula latifolia Lapeyrouse

3015 Coenosia sp. Meigen 1826 879 Sempervivum montanum L.

3016 Coenosiinae subfam. 880 Sempervivum arachnoideum L. 3017 Colletes impunctatus Nylander 1852 882 Sempervivum tectorum ssp. Alpinum L.

3018 Coreus sp. Fabricius 1794 887 Saxifraga paniculata Miller

3019 Cryptocephalus aureolus Suffrian 1847 901 Saxifraga aizoides L. 3020 Dasysyrphus friuliensis van der Goot 1960 909 Saxifraga bryoides L.

3021 Dilophus femoratus Meigen 1804 910 Saxifraga aspera L.

3022 Empididae sp. 936 Geum reptans L. 3023 Empis sp. L. 1758 969 Potentilla aurea L.

3024 Empria fletcheri Cameron 1878 1129 Trifolium pratense ssp. pratense L.

3025 Episyrphus balteatus De Geer 1776 1135 Trifolium pallescens Schreber 3026 Eristalis tenax L. 1758 1145 Trifolium badium Schreber

3027 Eupeodes corollae Fabricius 1794 1150 Anthyllis vulneraria ssp. alpestris Schultes

3028 Eupeodes latilunulatus Collin 1931 1158 Lotus alpinus Ramond

3029 Eupeodes nitens Zetterstedt 1843 1268 Epilobium fleischeri Hochstetter

3030 Halictus rubicundus Christ 1791 1297 Thesium alpinum L.

3031 Helina annosa Zetterstedt 1838 1499 Laserpitium halleri Crantz

3032 Hybomitra sp. Enderlein 1922 1585 Myosotis scorpioides L.

3033 Hybotidae sp. Enderlein 1922 1701 Thymus sp.

3034 Exochus sp. Gravenhorst 1829 1761 Linaria alpina ssp. alpina (L.) Miller 3035 Lasioglossum alpigenum Dalla Torre 1877 1796 Veronica fructicans Jacques

3036 Lasioglossum fratellum Perez 1903 1824 Euphrasia minima Schleicher

3037 Leucozona lucorum L. 1758 1846 Rhiantus minor L. 3038 Malthodes sp. Kiesenwetter 1852 1853 Melampyrum silvaticum L.

3039 Meliscaeva auricollis Meigen 1822 1900 Campanula barbata L.

3040 Miridae sp. 1908 Campanula cochleariifolia Lamarck

3041 Muscidae sp. 1917 Phyteuma hemisphaericum L.

3042 Myopa sp. Fabricius 1775 1923 Phyteuma betonicifolium Villars

3043 Nematus bohemani Thomson 1871 1954 Galium anisophyllon Villars

3044 Osmia inermis Zetterstedt 1838 2012 Valeriana tripteris L.

3045 Osmia loti Morawitz 1867 2028 Adenostyles leucophylla (Willd.) Reichenbach

3046 Paragus punctulatus Zetterstedt 1838 2030 Solidago virgaurea ssp. minuta (L.) Archangeli 3047 Phaonia hybrida Schnabl 1888 2051 Erigeron acer ssp. politus (Fr.) Lindberg

3048 Phaonia meigeni Schnabl 1888 2067 Anthennaria dioica (L.) Gaertner

3049 Phaonia serva Meigen 1826 2114 Achillea erba-rotta ssp. moschata (Wulfen) Vacc.


3050 Phaonia sp. Robineau-Desvoidy 1830 2139 Leucanthemopsis alpina (L.) Heywood

3051 Phoridae sp. Robineau-Desvoidy 1830 2271 Leontodon helveticus Mérat 3052 Phratora vitellinae L. 1758 2297 Taraxacum sp.

3053 Pipiza bimaculata Meigen 1822 2343 Hieracium staticifolium Allioni

3054 Pipizella sp. Rondani 1856 2357 Hieracium murorum L. 3055 Platycheirus albimanus Fabricius 1781

3056 Platycheirus melanopsis Loew 1856

3057 Platycheirus scutatus Meigen 1822 3058 Platypalpus sp. Le Peletier & Serville 1828

3059 Psilidae sp.

3060 Rhamphomyia sp. Meigen 1822 3061 Rhingia campestris Meigen 1822

3062 Sarcophagidae sp.

3063 Scaeva pyrastri L. 1758 3064 Scaeva selenitica L. 1758

3065 Sciaridae sp.

3066 Siphona sp. Meigen 1803 3067 Sphaerophoria bankowskae Goeldlin 1989

3068 Sphaerophoria laurae Goeldlin 1989

3069 Sphaerophoria scripta L. 1758 3070 Sphaerophoria sp. Le Peletier & Serville 1828

3071 Ichneumon sp. L. 1758

3072 Spilogona sp. Schnabl 1911 3073 Stomoxys calcitrans L. 1758

3074 Syrphus ribesii L. 1758

3075 Syrphus torvus Osten-Sacken 1875 3076 Syrphus vitripennis Meigen 1822

3077 Tachinidae sp.

3078 Tenthredo brevicornis Konow 1886 3079 Thricops aculeipes Zetterstedt 1838

3080 Thricops cunctans Meigen 1826

3081 Thricops longipes Zetterstedt 1845 3082 Thricops nigritellus Zetterstedt 1838

3083 Thricops rostratus Meade 1882

3084 Thricops semicinereus Wiedemann 1817 3085 Thricops separ Zetterstedt 1845

3086 Thricops sp. Zetterstedt 1845

3087 Thricops sudeticus Schnabl 1888 3088 Pipizella annulata Macquart 1829

3089 Bombus pratorum L. 1761

3093 Cryptinae subfam. 3094 Tryphoninae subfam.

3095 Campopleginae subfam.

803

804


805

806

807

808

809

810

T1 T2 Site T1 T2 Site T1 T2 Site T1 T2 Site T1 T2 Site T1 T2 Site T1 T2 Site

Qualitative network properties

Network size 17 15 25 16 18 27 31 27 42 24 24 37 33 37 54 26 NA 26 41 24 49Number of visited plant species 4 3 4 4 3 5 9 10 12 11 11 16 12 9 15 17 NA 17 26 14 20Number of not visited plant species 12 8 13 10 7 11 15 7 13 9 12 12 12 6 10 2 NA 2 17 11 3Number of pollinator species 13 12 21 12 15 22 22 17 30 13 13 21 21 28 39 9 NA 9 15 10 29Links per species 0.77 0.87 0.92 0.75 0.83 0.85 0.94 0.78 1.02 0.75 0.71 0.89 0.94 0.92 1.11 1.08 NA 1.08 1.07 0.96 1.2Connectance 0.25 0.36 0.27 0.25 0.33 0.21 0.15 0.12 0.12 0.13 0.12 0.10 0.12 0.13 0.10 0.18 NA 0.18 0.11 0.16 0.10Average degree 1.53 1.73 1.84 1.5 1.67 1.7 1.87 1.56 2.05 1.5 1.42 1.78 1.88 1.84 2.22 2.25 NA 2.25 2.15 1.92 2.41Plant degree 3.25 4.33 5.75 3 5 4.6 3.22 2.1 3.58 1.64 1.55 2.06 2.58 3.78 4 3.11 NA 3.11 2.93 2.3 2.95Pollinator degree 1 1.08 1.1 1 1 1.05 1.32 1.24 1.43 1.38 1.31 1.57 1.48 1.21 1.54 1.65 NA 1.65 1.69 1.64 2.03

130

Site age (years)

49 63 84 1098 28

Appendix 2. Qualitative (unweighted) network properties of plant–pollinator networks at seven sites of increasing distance from the glacier

snout across the Morteratsch glacier foreland, representing a chronosequence from 8 to 130 years since deglaciation. T1 = value for the 50 m

long transect; T2 value for the 30 m long transect; Site = value for the pooled data of the two transect for a site. At the 109-year old site only one

transect could be sampled. For the calculation of the network properties only the number of flowering plant species visited by pollinators were

used.

Albrecht ________

et al. Plant–pollinator network assembly 41 _______________________________________________________________

811

812

813

814

815

816

817

818

Site pair Modified Simpson's index

8 – 28 0.308 – 49 0.228 – 63 0.008 – 84 0.13

8 – 109 0.048 – 130 0.0028 – 49 0.1728 – 63 0.0928 – 84 0.17

28 – 109 0.1328 – 130 0.0449 – 63 0.1249 – 84 0.16

49 – 109 0.2049 – 130 0.0263 – 84 0.24

63 – 109 0.1263 – 130 0.1284 – 109 0.1784 – 130 0.08

109 – 130 0.14

Appendix 3. Values of the modified Simpson’s index as a measure of interaction

similarity of the plant–pollinator networks for each site pair. Plant–pollinator

interactions were sampled at seven sites of increasing distance from the glacier snout

across the Morteratsch glacier foreland, representing a chronosequence from 8 to 130

years since deglaciation. The numbers in the site pair column indicate the ages of the

two sites in years.

Documents

University of Zurich · predictions of increasing specialization of pollination systems during succession, but are in agreement with expectations from optimal foraging theory, predicting