www.sciencemag.org/cgi/content/full/335/6075/1486/DC1

Supporting Online Material for

Specialization and Rarity Predict Nonrandom Loss of Interactions from Mutualist Networks

Marcelo A. Aizen,* Malena Sabatino, Jason M. Tylianakis

*To whom correspondence should be addressed. E-mail: maizen@comahue-conicet.gob.ar

Published 23 March 2012, Science 335, 1486 (2012)

DOI: 10.1126/science.1215320

This PDF file includes

Materials and Methods SOM Text Figs. S1 to S6 Tables S1 to S4 Full References

  

2  

Materials and Methods

Study system

The 12 sierras surveyed in this study were located between Mar del Plata (37o58' S, 57o35' W) and Balcarce (37o50' S, 58o15' W), Buenos Aires Province, Argentina, which lie in the southeast of the ‘‘Pampeano Austral’’ biogeographic district (36). These sierras are part of the Tandilian orographic system (lower Paleozoic), which comprises about 24 isolated hills up to ~500 m in altitude. The study sierras ranged in area from 13 to 2100 ha (table S1), and each sierra was located between 0.8 and 9.1 km from its nearest neighbor. This range of area and isolation is typical of the entire orographic system. Sierra size was uncorrelated with proximity to other sierras, reducing the likelihood of spatial autocorrelation confounding results (13). The region has a temperate climate, with warm summers (mean January temperature 20.8oC) and mild winters (mean July temperature 5.2oC).

The sierras typically include three main habitats: a gentle rocky basal slope dominated by diverse shrubs, herbs, and geophytes; a barely-vegetated steep scarp; and a flat top with a mosaic of exposed bedrock and loessic patches dominated by grasses. The most abundant insect-pollinated plant families present in the sierras are the Asteraceae, Apiaceae, Fabaceae, and Scrophulariaceae, whereas the flower-visitor assemblage predominantly comprises Hymenoptera, followed by Diptera, Coleoptera, and Lepidoptera (13). The sierras are nested within an intensively-managed agricultural matrix, dominated by soybean, sunflower, wheat, corn, potato and canola, in a landscape that was a natural grassland mosaic prior to cultivation. The sierras support a rich flower-visitor community (13), despite intensive apiculture with European honey bees (Apis mellifera) in the surrounding matrix. We conducted fieldwork during the 2007–2008 flowering season (October–April). On each sierra, we sampled 0.5 ha on a north-facing rocky slope about 200 m from the edge of the nearest agricultural field. These sun-exposed slopes exhibit the highest plant diversity among the sierra habitats. Within each 0.5 ha area, we established two parallel 100-m transects, 50 m apart, along which we set 10 permanent 1-m radius plots, five plots per transect, approximately 25 m apart. We identified all blooming angiosperm species within each plot, and identified and counted all flower visitors that contacted floral sexual organs during a 15-min period. All plots from a given sierra were sampled consecutively between 09:00 and 18:00 h. We conducted observations on two sierras per day, and each sierra was sampled an average of 10 times throughout the flowering season, at two-week intervals. In total, we accumulated 318 h of observations, distributed over 127 days. Plant species that received no pollinator visits were excluded from analysis. Individual animals observed to visit flowers were morphotyped and identified with the aid of a reference collection and specialist help. Flower visitors (hereafter "pollinators") that could be identified to species or genus accounted for 52 or 80% of all individuals, respectively. For simplicity, here we refer to all morphospecies as species.

Data analysis

Our first hypothesis was that the loss of interactions with decreasing sierra size was non-random, i.e. that certain interactions would be progressively lost from the network. To test this, we analyzed interaction “nestedness”, i.e. the extent to which plant-pollinator interactions (and also plant/animal species) in a smaller sierra comprised a proper subset of

  

3  

interactions (or species) that were present in the next largest sierra. Presence and absence of interactions (and species) across the sierras was summarized in a binary matrix, with rows representing interactions (or plant or animal species) and columns identifying sierras. The nestedness of these matrices was then tested using the Nestedness metric based on Overlap and Decreasing Fill (NODF) of Almeida-Neto et al. (20), implemented in the ANINHADO package (37). We selected NODF from among the many nestedness indices (38), because it allows the separate estimation of nestedness among rows or columns in addition to matrix-wide; and among either rows or columns arrayed by criteria other than decreasing marginal sums (20), such as sierra size. We calculated NODF for the presences of plant-pollinator interactions, plant species, and pollinator species among sierras (i.e., columns) ordered by decreasing sierra size, as this relates directly to the inferred gradient of habitat reduction. Each observed index was compared with a distribution of expected indices based on 1000 randomly-generated matrices according to two null models: (i) re-assignment of presences with equal probability to any cell within the matrix, and (ii) re-assignment of presences to a given cell proportional to the average of the respective row and column marginal probabilities (20). The first model preserves the total number of occurrences, but otherwise randomizes the original structure of the meta-community, whereas the second model incorporates some original structure by distinguishing among widespread and restricted interactions or species, and between sierras rich or poor in interactions or species. Observed NODFs (X´s) were Z-transformed, /SD, where and SD are the mean and standard deviation, respectively, of a randomly-generated distribution. Z-scores allow both testing of observed nestedness against an expected normal probability density distribution, with mean = 0 and SD = 1, and comparison of observed nestedness among binary matrices of different shape and completeness (20).

After establishing that interactions were lost non-randomly, we aimed to determine whether any traits of an interaction predisposed it to be lost more rapidly from the network. The two traits we examined were interaction frequency (how commonly the interaction between a given plant and pollinator was observed in a given sierra) and degree of generalization (how dependent the interacting species pair was on that specific interaction). We asked, for each sierra, whether these two traits of an interaction explained its ubiquity (the most ubiquitous interactions being those observed in the most sierras). For each plant-pollinator interaction in each pollination web (i.e., focal sierra), we defined ubiquity as the proportion of all other sampled sierras in which that interaction was observed, excluding the focal sierra (i.e., from 0 to 11, divided by 11). Interaction frequency was estimated as the (log10) number of individual flower visitors to a particular plant species observed during the study at a focal sierra. Thus, at each sierra we pooled observations from the ten permanent plots over all sampling dates. The degree of generalization of an interaction was defined as the average number of species with which each species involved in an interaction interacted at a focal sierra. For example, in the most specialized (i.e., least generalized) interactions, the two partners (plant and pollinator) interacted only with each other (i.e., degree of generalization = 1). To factor out influences of web size and connectance on these traits and to make results among sierras comparable, we standardized interaction frequency and degree of generalization according to / , where and  are the minimum and maximum values among all interactions recorded at a given sierra. Thus, for each sierra, relative (i.e., standardized) interaction frequency and degree of generalization varied between 0, for the most infrequent or specialized interaction, and 1, for the most frequent and generalized interaction.

  

4  

For each sierra, we quantified the strength of the relationship between interaction ubiquity and its predictors (interaction frequency and degree of generalization) using the regression coefficients from a binomial generalized linear model (with a logit link function) (39). To account for any potential collinearity between predictors, we repeated these analyses using partial regression coefficients (i.e., the independent effect of each predictor, holding the effect of the other predictor constant) from a model that included both predictors (traits). Because of overdispersion, we fitted our models using quasi-likelihood (39).

Our second hypothesis was that the strength of this relationship between the ubiquity of an interaction and its traits (i.e., interaction frequency or degree of generalization) within sierras (fig. S1) would weaken with sierra size. The rationale for this was that networks in small sierras should have been already depleted of their most susceptible interactions, with only the abundant, ubiquitous interactions remaining plus some less-ubiquitous, perhaps facultative interactions that might occur at any frequency and have any degree of generalization. Thus, to determine whether the dependence of interaction ubiquity on interaction frequency and degree of generalization varied with sierra size, we regressed the simple and partial regression coefficients above against sierra size. All linear models were conducted using the glm and lm functions in the base package of R v. 2.10.1 (40).

Although there was no association between sierra size and proximity to other sierras (13), it is theoretically possible that the results we present here could be confounded by spatial autocorrelation among the sierras (for example, if a species with a disproportionate effect on network structure could only disperse among sites that were close together). To control for any such possibilities, we re-ran the linear models, which tested whether the strength of the trait-ubiquity relationships changed with sierra size, using a simultaneous autoregression conducted in SAM v.4.0 (41). This form of regression model explicitly incorporates the spatial relationship between pairs of sites, and can thereby test for a relationship between two or more variables after controlling for spatial proximity (42). We found that the relationships presented here (Fig. 2) were robust after controlling for spatial association (table S3).

Also, weakening of the relationships between ubiquity and the two interaction traits, frequency and degree of generalization, as sierra size decreases, could simply reflect lower variability of the predictors (i.e., before standardization) in the smallest sierras. However, the coefficient of variation (CV) of interaction frequency varied inversely with (log10) sierra size (F1,10 = 6.34, P = 0.03; table S2), whereas the CV of interaction degree showed no trend (F1,10 = 0.24, P = 0.64). Thus, the relationships reported in Fig.2 were not artifacts of positive size-correlated changes in the range of values of the two predictors.

Finally, our results suggested that many interactions forming the core of the network in larger sierras were present but were not part of the core in smaller sierras. Therefore, for these ubiquitous interactions (defined here as those occurring at least in six of the 12 sierras) we assessed the hypothesis that they were displaced from the web core (i.e., they demonstrated a loss of centrality) with decreasing sierra size. Loss of centrality would imply that these interactions show a decrease in frequency and degree of generalization, within each sierra, as sierra size decreases. To test this, we ran two general linear mixed models (39), where we considered (log10) sierra size as a predictor of relative (i.e., standardized) interaction frequency or degree of generalization. Interaction and sierra identity were both included as random, grouping factors (to determine how sierra size

  

5  

affected, on average, each given interaction, and to control for the non-independence of interactions within a sierra). Although for these tests we considered ubiquitous interactions to be those occurring in at least 50% of the sierras (68 out of a total of 1170 interactions), the significance and strength of the results of this analysis (fig. S6) were not affected by the use of a more stringent definition of ubiquity (i.e., those interactions present in 7, 8, 9, 10, 11 or 12 sierras). These hierarchical models were conducted using the lme4 package (43) in the R environment v. 2.10.1 (40). 

SOM Text

Missing interactions

Mutualistic interactions are temporally variable, and "missing interactions" could appear during some other year or within a season at times when the observer was not in the field (44). However, the problem of interactions not being sampled is common to all empirical food-web and mutualist-network studies (e.g., the datasets used in 26, 29, 45). A few studies have attempted to quantify the turnover of individual network interactions through time (46, 47), and some papers have suggested the use of rarefaction curves to estimate the number of missing links (48, 49). However, it is not yet clear whether rarefaction curves for interactions can be interpreted in the same way as for species. For instance, the finding that four years of sampling only detected 50% of estimated interaction diversity in a desert pollination network suggests that asymptotic diversity estimators may actually over-estimate the number of interaction links (48). It is known, however, that some emergent properties of mutualistic networks, such as nestedness and connectance, seem to asymptote quickly with increasing sampling intensity, and are therefore less affected by web sampling effort (16, 46, 50).   Here, we deliberately standardized our sampling effort (i.e., time spent, area surveyed, and habitat type sampled) across sierras to make inferences from our comparisons valid. In addition, the total number of individual flower visitors sampled was unrelated to sierra size (F1,10 = 0.02, P = 0.89; both variables log-transformed), so that none of the patterns we report can be attributed to any artifact related to area-correlated differences in sample size. Finally, we found that the effect of the degree of generalization on interaction ubiquity persisted after accounting for differences in interaction frequency (fig. S4), which suggests that our results are robust and not confounded by the presence of undersampled interactions. Therefore, the main point in our manuscript is that, within the limitations of any realistic field sampling campaign and beyond a background of "random noise" determined by sampling itself, we were able to infer a deterministic pattern of interaction erosion that we could associate with specific interaction characteristics. This does not mean that some of the "missing" interactions could not have been sampled in a small sierra if we would be able to increase our sampling effort over time and space. However, we can generalize from our results that small habitat remnants have a deficit of vulnerable mutualistic interactions, identified here as those occurring naturally at low frequency and between species with a low number of alternative interaction partners. 

A

B

C

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Vigilancia (2147 ha)

Relative interaction frequency

Ubi

quity

t=7.39, P<0.000

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=7.06, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Volcan (2079 ha)

Relative interaction frequency

Ubi

quity

t=9.13, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=7.74, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Difuntos (1692 ha)

Relative interaction frequency

Ubi

quity

t=6.37, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=5.71, P<0.0001

Difuntos (1692 ha)

Animals

Pla

nts

9588726058472312387368108094891635783359399

44322

105

106

118

108

110

101

223

249

103

125

188

197

222

259

115

192

204

220

234

262

264

265 99 109

134

154

165

203

214

117

121

133

150

164

186

211

218

221

233

257

263

111

113

114

123

126

127

129

132

136

137

145

147

148

163

167

168

175

179

189

198

200

201

210

215

224

238

244

247

0

2

4

6

8

10

12

Volcan (2079 ha)

Animals

Pla

nts

9492908367611913896562595136211545807538449529

816

2393330

778

93732

108

109

121

223

234

101

106

125

103

105

118

133

165

222

203

262

210

248

259

111

115

160

188

200

201

239

265

117

131

137

154

163

164

181

192

197

204

211

218

263 98 113

116

124

126

128

130

132

135

138

144

145

150

152

162

172

173

175

194

199

202

219

231

233

238

244

257

264

0

2

4

6

8

10

12

Vigilancia (2147 ha)

Animals

Pla

nts

93685851434024133

85805919147974658

767

44185595371629339

7821302

32

108

105

106

259

103

115

223

262

125

165

210

101

121

164

222

109

118

133

154

163

218

234

265

111

233

119

131

137

156

188

204

238 99 110

129

132

140

147

150

175

191

203

211

214

220

246

248

263 97 112

116

122

142

151

159

162

166

168

172

173

177

192

197

198

205

209

212

217

239

244

250

257

260

261

0

2

4

6

8

10

12

D

E

F

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

La Brava (396 ha)

Relative interaction frequency

Ubi

quity

t=5.32, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

La Barrosa (243 ha)

Relative interaction frequency

Ubi

quity

t=6.56, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Cinco Cerros (541 ha)

Relative interaction frequency

Ubi

quity

t=5.98, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=4.49, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=5.01, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=7.28, P<0.0001

La Barrosa (243 ha)

Animals

Pla

nts

757365584741382015863

9376715128258987136356463685955955447

379

782

32

108

105

106

110

259

262

121

177

101

109

115

210

103

118

133

165

192

201

218

220

222

234

263

267 99 125

137

145

150

173

188

197

200

238

111

113

127

128

132

143

146

149

154

168

170

175

196

199

204

214

216

223

249

250

257

260

264

268

0

2

4

6

8

10

12

La Brava (396 ha)

Animals

Pla

nts

9073679480755136211585583830957898

3972

374432

108

121

103

101

203

222

165

118

173

223

259 99 105

125

188

197

262

106

160

210

218

238

247

115

116

131

133

137

145

150

199

200

234

244

257

263 97 109

110

112

129

146

149

154

163

164

169

175

180

192

198

204

229

240

250

256

265

268

0

2

4

6

8

10

12

Cinco Cerros (541 ha)

Animals

Pla

nts

908676757358545150353327786563533695595538472

669

77374430567

2932

108

105

106

259

121

125

133

218

223

234

262

265

103

137

192

101

109

116

145

157

164

204

210

220

112

131

155

163

173

188

201

213

244

263

264

102

110

113

115

117

118

119

132

141

146

154

158

160

165

166

182

190

202

205

211

221

222

232

233

238

241

242

254

0

2

4

6

8

10

12

G

H

I

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

La Chata (229 ha)

Relative interaction frequency

Ubi

quity

t=7.34, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=7.04, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Amarante (190 ha)

Relative interaction frequency

Ubi

quity

t=6.08, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=4.96, P<0.0001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

El Morro (49 ha)

Relative interaction frequency

Ubi

quity

t=2.24, P=0.027

El Morro (49 ha)

Animals

Pla

nts

7346148765502

5556513859788

763234

106

108

234

101

103

259

105

173

192

210

262

121

125

131

145

165

218

220

113

117

133

137

150

172

188

204

208

214

248

260

265

109

115

122

128

132

134

139

152

153

154

155

160

162

164

170

171

174

177

185

202

203

223

224

225

232

235

236

237

240

244

245

246

247

251

257

263

0

2

4

6

8

10

12

Amarante (190 ha)

Animals

Pla

nts

96948375736650494736312813867811

4916559517655427785953830

87

8092

9045375632

108

106

121

103

259

234

109

116

188

137

105

125

133

210

101

115

160

195

211

218

262

263

265

145

154

165

173

175

192

201

203

214

220

238

258 99 107

111

113

117

122

126

130

131

132

139

143

144

146

148

150

152

157

159

162

167

168

186

197

204

213

222

223

241

249

250

257

0

2

4

6

8

10

12

La Chata (229 ha)

Animals

Pla

nts

91696552512821

394787336

8893824

97

75639559484593907637164430

232

108

106

105

125

131

259

218

222

101

177

188

210

223

133

103

109

118

132

137

150

165

197

200

201

234

262

263

265

110

115

116

121

128

145

154

163

173

192

203

204

214

217

256

107

119

120

122

140

144

170

185

196

198

220

238

240

249

250

255

257

264

0

2

4

6

8

10

12

0.0 0.2 0.4 0.6 0.8 1.0

0.00.20.40.60.81.0

Relative interaction degree

Ubi

quity

t=-1.63, P=0.106

J

K

L

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Piedra Alta (13 ha)

Relative interaction frequency

Ubi

quity

t=2.88, P=0.004

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=-2.19, P=0.030

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Difuntito (13 ha)

Relative interaction frequency

Ubi

quity

t=2.91, P=0.004

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

La Paja (12 ha)

Relative interaction frequency

Ubi

quity t=3.64, P<0.001

0.0 0.2 0.4 0.6 0.8 1.0

0.0

0.2

0.4

0.6

0.8

1.0

Relative interaction degree

Ubi

quity

t=4.44, P<0.0001

La Paja (12 ha)

Animals

Pla

nts

8380726954502817

541

55897359513795

7786590

25638

97632

108

106

109

105

210

259

103

111

117

262

121

137

154

188

220

223

101

115

118

152

160

165

171

213

218

222

234

251

265

122

128

131

133

145

146

150

173

175

178

179

192

204

211

225

232

233

244

250

0

2

4

6

8

10

12

Piedra Alta (13 ha)

Animals

Pla

nts

8137923690856551232011737228302

593255384713

108

105

106

125

131

262

173

137

101

109

163

188

210

259

265

118

134

154

164

218

234

100

103

107

110

115

116

117

121

122

147

165

169

192

213

252

263

266

104

119

123

126

148

160

161

162

167

172

177

183

185

186

197

206

207

208

211

214

222

224

225

232

233

243

249

257

0

2

4

6

8

10

12

Difuntito (13 ha)

Animals

Pla

nts

898482777370572217137

926967594726642

665535281

68509065118678328

7638303756

108

259

106

101

109

116

126

234 99 103

162

223

262

121

133

173

137

218

227

244

264

265

105

118

119

147

251

257

100

112

117

145

154

155

165

184

186

188

192

213

214

220

232

246

249

260

263

267

107

110

115

122

131

132

148

160

163

164

175

176

178

187

193

197

203

204

206

208

210

211

222

226

228

229

230

247

253

254

266

0

2

4

6

8

10

12

0.0 0.2 0.4 0.6 0.8 1.0

0.00.20.40.60.81.0

Relative interaction degree

Ubi

quity

t=0.52, P=0.601

  

10  

Fig. S1. The plant-pollinator interaction matrix, and the relationships between interaction ubiquity and its two predictors, interaction frequency and degree of generalization, for each of the 12 sierras (A to L). Sierras are ordered by decreasing size. For the matrices, a colored cell specifies an observed interaction and different colors and color hues indicate the total number of sierras in which each interaction was found (from 1 to 12). In each matrix, species are ranked according to decreasing number of interactions per species for the corresponding sierra. Species/morphospecies names related to each number code are listed in table S4. For the right two panels, symbol area is proportional to the number of overlapping individual observations. Relative (i.e., standardized) interaction frequency and degree of generalization were calculated according to  / . Solid lines illustrate best-fit binomial linear models.

  

11  

Fig. S2. Z-transformed Nestedness metric based on Overlap and Decreasing Fill (NODF) of pollination interactions (I), and plant (P) and animal (A) species involved in the pollination networks of 12 sierras ordered by decreasing area, based on two alternate null models. Z > 1.96 or < -1.96 (dotted lines) indicates significantly (P < 0.05) greater or lower nestedness, respectively, than null expectations based on (A) completely random assignment of presences to any cell within the matrix and (B) assignment of presences to any cell in proportion to the average of its row and column probabilities (20).

  

12  

Fig. S3. Pearson´s correlations (r + 95% CI´s) between interaction frequency and degree of generalization for each of the 12 sierras regressed against sierra area. Individual correlations whose confidence intervals do not overlap the dotted line differ significantly from zero. Solid lines and summary statistics indicate that the relationship between the two interaction traits increased marginally with sierra area.

   

  

13  

 

Fig. S4.  The dependence on sierra size of the relationship between interaction ubiquity and interaction (A) frequency and (B) degree of generalization. Dependence is represented by the partial regression coefficients (βpartial + 95% CI´s) from binomial generalized linear models conducted for each of the 12 sierras, which account for any collinearity between the predictors. Individual coefficients whose respective confidence intervals do not overlap the dotted line differ significantly from zero. Solid lines and summary statistics indicate that the relationships between ubiquity and each interaction trait increase significantly with sierra area.   

  

14  

 

 

 

 

Fig. S5.  Spearman´s correlations (r + 95% CI´s) between plant and animal specialization (i.e., degree) in each of the 12 sierras regressed against sierra area. Solid lines and summary statistics indicate that the association between plant and animal degree of generalization becomes, on average, less negative with increasing sierra size, i.e. interactions become more asymmetric in smaller sierras.   

  

15  

 

 

 

 

Fig. S6.  The dependence on sierra size of the relative (i.e., standardized) interaction (A) frequency and (B) degree of generalization shown by the ubiquitous interactions (i.e., interactions occurring in at least 50% of the surveyed sierras). Values are adjusted means + SEM based on general linear mixed models that considered (log10) sierra size as a predictor of relative interaction frequency or degree of generalization. Interaction and sierra identity were both included as random, grouping factors. Solid lines and summary statistics indicate that these ubiquitous interactions increase significantly their frequency of interaction and degree of generalization with sierra area.

  

16  

Table S1. Geographical coordinates and characteristics of the 12 plant–pollinator webs (i.e., sierras) analyzed. Sierras are ordered from the largest to the smallest.

Sierra Latitude Longitude Area (ha)

Number of plant

species

Number of animal species

Number of interaction

links

Number of flower visitors

Vigilancia -37.880881 -58.012314 2147 34 74 243 1167

Volcan -37.855217 -58.059364 2079 34 68 228 1311

Difuntos -37.885350 -57.841953 1692 25 69 192 1506

Cinco Cerros -37.737917 -58.240992 541 33 63 171 944

La Brava -37.881586 -57.986611 396 24 57 171 760

La Barrosa -37.873775 -58.263144 243 36 58 175 842

La Chata -37.877511 -58.384603 229 33 61 191 1101

Amarante -37.845356 -58.365892 190 39 67 188 889

El Morro -37.739264 -58.424703 49 17 67 138 825

Difuntito -37.760303 -58.250575 13 38 79 241 799

Piedra Alta -37.733280 -58.311856 13 22 66 168 1450

La Paja -37.751733 -58.289511 12 28 48 132 1581

  

17  

Table S2. Basic descriptive statistics of the two predictive interaction traits, interaction frequency and degree of generalization before standardization (see Data analysis in Materials and Methods). The values provided are the mean, standard deviation (SD), and coefficient of variation (CV). Sierras are ordered from the largest to the smallest.

Sierra Area (ha) Number of

interaction links

Interaction frequency Interaction degree

Mean SD CV (%) Mean SD CV (%)

Vigilancia 2147 243 0.388 0.434 112.0 10.69 5.326 49.8

Volcan 2079 228 0.393 0.472 120.1 10.50 4.959 47.2

Difuntos 1692 192 0.408 0.488 119.5 10.94 5.466 50.0

Cinco Cerros 541 171 0.392 0.465 118.6 8.15 3.610 44.3

La Brava 396 171 0.288 0.409 141.9 9.79 4.489 45.9

La Barrosa 243 175 0.367 0.459 125.0 9.05 4.335 47.9

La Chata 229 191 0.366 0.482 131.8 10.58 5.549 52.4

Amarante 190 188 0.315 0.435 137.9 7.88 3.707 47.0

El Morro 49 138 0.418 0.495 118.4 13.01 6.931 53.3

Difuntito 13 241 0.287 0.372 129.5 9.98 4.685 47.0

Piedra Alta 13 168 0.314 0.494 157.4 9.05 4.502 49.8

La Paja 12 132 0.430 0.593 137.9 8.45 4.117 48.7

  

18  

Table S3. Results of simultaneous autoregressions (SAR´s) testing for a relationship between (log-transformed) area of the focal sierra and (A) interaction frequency (Fig. 2A) and (B) interaction degree of generalization (Fig. 2B). In both cases, addition of spatial association to the model improved fit (reduced AICc and increased r2) compared with ordinary least squares (OLS), but the effect of area remained significant after controlling for spatial autocorrelation.

A. Interaction frequency

OLS Result: r: 0.876 r²: 0.767 AICc: 21.769

Explained by Predictor Variables: r: 0.875 r²: 0.765 AICc: 21.863

Total Explained (Predictor + Space): r: 0.936 r²: 0.877 AICc: 14.097

Variable OLS Coeff. SAR Coeff. Std Coeff. Std Error t P

Constant 0.794 0.673 0 0.639 1.053 0.32

log10(area) 0.724 0.759 0.918 0.158 4.806 <0.001

B. Interaction degree

OLS Result: r: 0.667 r²: 0.446 AICc: 53.376

Explained by Predictor Variables: r: 0.656 r²: 0.431 AICc: 53.69

Total Explained (Predictor + Space): r: 0.818 r²: 0.669 AICc: 47.196

Variable OLS Coeff. SAR Coeff. Std Coeff. Std Error t P

Constant -1.05 -3.108 0 2.779 -1.118 0.292

log10(area) 1.335 1.577 0.789 0.687 2.295 0.047

   

  

19  

Table S4. Complete list of plant and animal species/morphospecies and number codes used in fig. S1.

Plants 40. Gamochaeta argentina 1. Abutilon terminale 41. Gamochaeta stachydifolia 2. Achyrocline satureoides 42. Gelasine azurea 3. Adesmia incana 43. Geranium dissectum 4. Alophia lahue 44. Gerardia genistifolia 5. Apodanthera sagitifolia 45. Glandularia dissecta 6. Arjona tuberosa 46. Gnaphalium cheiranthifolium 7. Baccharis articulata 47. Gomphrena perennis 8. Baccharis coridifolia 48. Grindelia buphthalmoides 9. Baccharis tandilensis 49. Gymnocalycium reductum

10. Bidens pilosa 50. Habranthus andersoni 11. Blumenbachia insignis 51. Habranthus gracilifolius 12. Canna glauca 52. Helianthemum brasiliense 13. Carduus acanthoides 53. Hypericum connatus 14. Cerastium mollissimum 54. Hypochaeris pampasica 15. Cleanthes brasiliensis 55. Hypochaeris rosengurtii 16. Colletia paradoxa 56. Hysterionica pinifolia 17. Commelina erecta 57. Lathyrus crassipes 18. Conium maculatum 58. Lathyrus nervosus 19. Convolvulus hermanniae 59. Lathyrus pubescens 20. Conyza bonariensis 60. Melilotus officinalis 21. Crysanthemum leucanthemum 61. Mimosa tandilensis 22. Cuphea fruticosa 62. Nothoscordum bonariense 23. Cynara cardunculus 63. Oenothera indecora 24. Cypella herbertii 64. Opuntia aff. elata 25. Cytisus monspessulanus 65. Oxalis articulata 26. Discaria longispina 66. Oxalis cordobensis 27. Dodonaea viscosa 67. Oxypetalum solanoides 28. Echium plantagineum 68. Passiflora caerulea 29. Eryngium floribundum 69. Pavonia cymbalaria 30. Eryngium horridum 70. Petunia axillaris 31. Eryngium nudicaule 71. Picris echioides 32. Eryngium stenophyllum 72. Rapistrum rugosum 33. Eupatorium bupleurifolium 73. Rodophiala bifidum 34. Eupatorium lanigerum 74. Rubus ulmifolius 35. Eupatorium macrocephalum 75. Senecio arechavaletae 36. Eupatorium squarrulosum 76. Senecio bravensis 37. Eupatorium subhastatum 77. Senecio grisebachii 38. Eupatorium tenacetifolium 78. Senecio madagascariensis 39. Foeniculum vulgare 79. Senecio pulcher

  

20  

 

80. Senecio selloi 118. Brachygastra lechiguana 81. Sida flavescens 119. Braconidae 1 82. Silene gallica 120. Braconidae 2 83. Sisyrinchium iridifolium 121. Camponotus sp. 84. Solanum chacoense 122. Campsomeris bistrimacula 85. Solidago chilensis 123. Campsomeris sp. 2 86. Sommerfeltia spinulosa 124. Cantharidae 1 87. Sonchus oleraceus 125. Cantharidae 2 88. Spartium junceum 126. Ceratina montana 89. Spilanthes stolonifera 127. Ceratina rupestris 90. Stevia satureiifolia 128. Chalcidoidea 1 91. Trixis stricta 129. Chalcidoidea 2 92. Turnera pinnatifida 130. Chalepogenus goeldianus 93. Valeriana polystachya 131. Chauliognathus scriptus 94. Verbena intermedia 132. Chauliognathus sp. 2 95. Vernonia echioides 133. Chilicola sp. 96. Wigginsia tephracantha 134. Chlorion sp.

135. Chloropidae Animals 136. Chlorostilbon aureoventris

97. Achyra similalis 137. Chrysodina sp. 98. Agathidinae 138. Chrysomelidae 1 99. Agraulis vanillae 139. Chrysomelidae 2

100. Agrotis sp. 140. Chrysomelidae 3 101. Allograpta sp. 141. Chrysomelidae 4 102. Alophophion sp. 142. Chrysomelidae 5 103. Anthomyiidae 143. Chrysopa sp. 104. Anthrax sp. 144. Cleridae 105. Apis mellifera 145. Cochliomyia macellaria 106. Astylus quadrilineatus 146. Coelioxys sp. 107. Audre epulus signata 147. Colias lesbia 108. Augochlora semiramis 148. Colletes cyaneus 109. Augochlorella ephyra 149. Colletes sp. 2 110. Augochloropsis erato 150. Colletes sp. 3 111. Augochloropsis sp. 2 151. Conoderus sp. 112. Augochloropsis sp. 3 152. Copestylum sp. 113. Bibionidae 153. Coreidae 114. Bombus atratus 154. Corticea immocerina 115. Bombus bellicosus 155. Culicidae 116. Bombyliidae 1 156. Curculionidae 1 117. Bombyliidae 2 157. Curculionidae 2

  

21  

158. Curculionidae 3 198. Megacyllene insignita 159. Curculionidae 4 199. Megacyllene spixi 160. Dactylozodes specie 200. Melissoptila tandilensis 161. Danaus plexipus erippus 201. Melyridae 162. Dialictus sp. 202. Microgastrinae 163. Discodon sp. 1 203. Mischocyttarus drewseni 164. Discodon sp. 2 204. Mordellidae 1 165. Drosophilidae 205. Mordellidae 2 166. Emesis russula 206. Muscidae 167. Entypus ferruginipennis 207. Mydidae 168. Entypus sp. 208. Neochrysis sp. 169. Epicauta sp. 209. Odonthocera flavicauda 170. Eristalinus taeniops 210. Palpada sp. 171. Eristalis tenax 211. Pampasatyrus gyrtone 172. Eudesmia australis 212. Panca subpunctuli 173. Euptoieta claudia hortensia 213. Panurgillus sp. 174. Formicidae 214. Parachytas sp. 175. Gaesischia sp. 215. Paramoecerus barbicornis 176. Geometridae 216. Paroxystoglossa sp. 177. Halictillus sp. 217. Pepsis sp. 178. Harmonia axyrides 218. Phalacridae 179. Hedriodiscus pulcher 219. Photinus fuscus 180. Heteroderes rufangulus 220. Platycheirus sp. 181. Ichneumonoidea 1 221. Podagritus sp. 182. Ichneumonoidea 2 222. Polistes sp. 183. Ichneumonoidea 3 223. Polybia scutellaris 184. Isepeolus viperinus 224. Pompilidae 185. Isodontia sp. 225. Prionyx sp. 186. Leioproctus indigoticus 226. Pseudagapostemon brasiliensis 187. Lema sp. 227. Pseudagapostemon pampeanus 188. Lerodea eufala 228. Pseudagapostemon singularis 189. Leucospis sp. 229. Pseudocentron sp. 190. Libythea carinenta 230. Pseudodoros sp. 191. Linepithema humile 231. Pyraloidea 192. Lucilia sericata 232. Pyrgus sp. 193. Megachile sp. 1 233. Ruizantheda divaricata 194. Megachile sp. 2 234. Sarcophagidae 195. Megachile sp. 3 235. Scelionidae 196. Megachile sp. 4 236. Sceliphron asiaticum 197. Megachilidae 237. Sphex sp.

22

References and Notes 1. U. Bastolla et al., The architecture of mutualistic networks minimizes competition and

increases biodiversity. Nature 458, 1018 (2009). doi:10.1038/nature07950 Medline

2. T. Okuyama, J. N. Holland, Network structural properties mediate the stability of mutualistic communities. Ecol. Lett. 11, 208 (2008). doi:10.1111/j.1461-0248.2007.01137.x Medline

3. N. Rooney, K. McCann, G. Gellner, J. C. Moore, Structural asymmetry and the stability of diverse food webs. Nature 442, 265 (2006). doi:10.1038/nature04887 Medline

4. C. N. Kaiser-Bunbury, S. Muff, J. Memmott, C. B. Müller, A. Caflisch, The robustness of pollination networks to the loss of species and interactions: A quantitative approach incorporating pollinator behaviour. Ecol. Lett. 13, 442 (2010). doi:10.1111/j.1461-0248.2009.01437.x Medline

5. L. P. Koh et al., Species coextinctions and the biodiversity crisis. Science 305, 1632 (2004). doi:10.1126/science.1101101 Medline

6. C. Fontaine, I. Dajoz, J. Meriguet, M. Loreau, Functional diversity of plant-pollinator interaction webs enhances the persistence of plant communities. PLoS Biol. 4, e1 (2006). doi:10.1371/journal.pbio.0040001 Medline

7. A. Pauw, Collapse of a pollination web in small conservation areas. Ecology 88, 1759 (2007). doi:10.1890/06-1383.1 Medline

8. N. M. Williams, C. Kremen, Resource distributions among habitats determine solitary bee offspring production in a mosaic landscape. Ecol. Appl. 17, 910 (2007). doi:10.1890/06-0269 Medline

9. A. Müller et al., Quantitative pollen requirements of solitary bees: Implications for bee conservation and the evolution of bee–flower relationships. Biol. Conserv. 130, 604 (2006). doi:10.1016/j.biocon.2006.01.023

10. W. Bond, Do mutualisms matter? Assessing the impact of pollinator and disperser disruption on plant extinction. Philos. Trans. R. Soc. Ser. B 344, 83 (1994). doi:10.1098/rstb.1994.0055

11. N. M. Williams et al., Ecological and life-history traits predict bee species responses to environmental disturbances. Biol. Conserv. 143, 2280 (2010). doi:10.1016/j.biocon.2010.03.024

12. J. M. Tylianakis, R. K. Didham, J. Bascompte, D. A. Wardle, Global change and species interactions in terrestrial ecosystems. Ecol. Lett. 11, 1351 (2008). doi:10.1111/j.1461-0248.2008.01250.x Medline

13. M. Sabatino, N. Maceira, M. A. Aizen, Direct effects of habitat area on interaction diversity in pollination webs. Ecol. Appl. 20, 1491 (2010). doi:10.1890/09-1626.1 Medline

23

14. M. A. Aizen, C. L. Morales, J. M. Morales, Invasive mutualists erode native pollination webs. PLoS Biol. 6, e31 (2008). doi:10.1371/journal.pbio.0060031 Medline

15. B. Padrón et al., Impact of alien plant invaders on pollination networks in two archipelagos. PLoS ONE 4, e6275 (2009). doi:10.1371/journal.pone.0006275 Medline

16. J. M. Tylianakis, E. Laliberté, A. Nielsen, J. Bascompte, Conservation of species interaction networks. Biol. Conserv. 143, 2270 (2010). doi:10.1016/j.biocon.2009.12.004

17. S. Lavorel, S. McIntyre, J. Landsberg, T. D. Forbes, Plant functional classifications: From general groups to specific groups based on response to disturbance. Trends Ecol. Evol. 12, 474 (1997). doi:10.1016/S0169-5347(97)01219-6 Medline

18. Materials and methods are available as supporting material on Science Online.

19. J. Crisci, S. Freire, G. Sancho, L. Katinas, Historical biogeography of the Asteraceae from Tandilia and Ventania mountain ranges (Buenos Aires, Argentina). Caldasia 23, 21 (2001).

20. M. Almeida-Neto, P. Guimaraes, P. R. Guimaraes, Jr., R. D. Loyola, W. Ulrich, A consistent metric for nestedness analysis in ecological systems: Reconciling concept and measurement. Oikos 117, 1227 (2008). doi:10.1111/j.0030-1299.2008.16644.x

21. D. Tilman, R. M. May, C. L. Lehman, M. A. Nowak, Habitat destruction and the extinction debt. Nature 371, 65 (1994). doi:10.1038/371065a0

22. D. P. Vázquez, M. A. Aizen, Null model analyses of specialization in plant–pollinator interactions. Ecology 84, 2493 (2003). doi:10.1890/02-0587

23. D. P. Vázquez, R. Poulin, B. R. Krasnov, G. I. Shenbrot, Species abundance and the distribution of specialization in host-parasite interaction networks. J. Anim. Ecol. 74, 946 (2005). doi:10.1111/j.1365-2656.2005.00992.x

24. J. Memmott, P. G. Craze, N. M. Waser, M. V. Price, Global warming and the disruption of plant-pollinator interactions. Ecol. Lett. 10, 710 (2007). doi:10.1111/j.1461-0248.2007.01061.x Medline

25. J. Bascompte, P. Jordano, C. J. Melián, J. M. Olesen, The nested assembly of plant-animal mutualistic networks. Proc. Natl. Acad. Sci. U.S.A. 100, 9383 (2003). doi:10.1073/pnas.1633576100 Medline

26. E. Thébault, C. Fontaine, Stability of ecological communities and the architecture of mutualistic and trophic networks. Science 329, 853 (2010). doi:10.1126/science.1188321 Medline

27. P. R. Guimarães Jr. et al., Interaction intimacy affects structure and coevolutionary dynamics in mutualistic networks. Curr. Biol. 17, 1797 (2007). doi:10.1016/j.cub.2007.09.059 Medline

24

28. J. N. Thompson, The Coevolutionary Process (Univ. of Chicago Press, Chicago, 1994).

29. J. Bascompte, P. Jordano, J. M. Olesen, Asymmetric coevolutionary networks facilitate biodiversity maintenance. Science 312, 431 (2006). doi:10.1126/science.1123412 Medline

30. D. P. Vázquez, M. A. Aizen, Asymmetric specialization: A pervasive feature of plant–pollinator interactions. Ecology 85, 1251 (2004). doi:10.1890/03-3112

31. J. M. Olesen, L. I. Eskildsen, S. Venkatasamy, Invasion of pollination networks on oceanic islands: Importance of invader complexes and endemic super generalists. Divers. Distrib. 8, 181 (2002). doi:10.1046/j.1472-4642.2002.00148.x

32. J. Memmott, N. M. Waser, M. V. Price, Tolerance of pollination networks to species extinctions. Proc. Biol. Sci. 271, 2605 (2004). doi:10.1098/rspb.2004.2909 Medline

33. B. Anderson, S. D. Johnson, The geographical mosaic of coevolution in a plant-pollinator mutualism. Evolution 62, 220 (2008). doi:10.1111/j.1558-5646.2007.00275.x Medline

34. K. E. Steiner, V. Whitehead, Pollinator adaptation to oil-secreting flowers--Rediviva and Diascia. Evolution 44, 1701 (1990). doi:10.2307/2409348

35. A. Pauw, J. A. Hawkins, Reconstruction of historical pollination rates reveals linked declines of pollinators and plants. Oikos 120, 344 (2011). doi:10.1111/j.1600-0706.2010.19039.x

36. A. L. Cabrera, A. Willink, Biogeografía de América latina (Organización de Estados Americanos, Washington DC, 1973).

37. P. R. Guimarães, Jr., P. Guimarães, Improving the analyses of nestedness for large sets of matrices. Environ. Model. Softw. 21, 1512 (2006). doi:10.1016/j.envsoft.2006.04.002

38. W. Ulrich, M. Almeida Neto, N. J. Gotelli, A consumer's guide to nestedness analysis. Oikos 118, 3 (2009). doi:10.1111/j.1600-0706.2008.17053.x

39. M. J. Crawley, The R Book (Wiley, West Sussex, 2007).

40. R Development Core Team, R: A language and environment for statistical computing (R Foundation for Statistical Computing, Vienna, Austria, 2011). http://CRAN.R-project.org

41. T. F. Rangel, J. A. F. Diniz Filho, L. M. Bini, SAM: A comprehensive application for spatial analysis in macroecology. Ecography 33, 46 (2010). doi:10.1111/j.1600-0587.2009.06299.x

42. C. F. Dormann et al., Methods to account for spatial autocorrelation in the analysis of species distributional data: A review. Ecography 30, 609 (2007). doi:10.1111/j.2007.0906-7590.05171.x

43. D. Bates, M. Maechler, lme4: Linear mixed-effects models using S4 classes, R package version 0.999375-35; http://CRAN.R-project.org/package=lme4.

25

44. P. C. de Ruiter, V. Wolters, J. C. Moore, K. O. Winemiller, Food web ecology: Playing Jenga and beyond. Science 309, 68 (2005). doi:10.1126/science.1096112 Medline

45. E. L. Rezende, J. E. Lavabre, P. R. Guimarães, P. Jordano, J. Bascompte, Non-random coextinctions in phylogenetically structured mutualistic networks. Nature 448, 925 (2007). doi:10.1038/nature05956 Medline

46. Y. L. Dupont, B. Padrón, J. M. Olesen, T. Petanidou, Spatio-temporal variation in the structure of pollination networks. Oikos 118, 1261 (2009). doi:10.1111/j.1600-0706.2009.17594.x

47. E. Laliberté, J. M. Tylianakis, Deforestation homogenizes tropical parasitoid-host networks. Ecology 91, 1740 (2010). doi:10.1890/09-1328.1 Medline

48. N. P. Chacoff et al., Evaluating sampling completeness in a desert plant-pollinator network. J. Anim. Ecol. 81, 190 (2012). doi:10.1111/j.1365-2656.2011.01883.x Medline

49. J. M. Olesen et al., Missing and forbidden links in mutualistic networks. Proc. R. Soc. B. Biol. 278, 725 (2011). doi:10.1098/rspb.2010.1371 Medline

50. A. Nielsen, J. Bascompte, Ecological networks, nestedness and sampling effort. J. Ecol. 95, 1134 (2007). doi:10.1111/j.1365-2745.2007.01271.x