1

Patterns of migratory connectivity in Vaux’s Swifts at a northern migratory roost: a multi-

2

isotope approach

3

Running head: Vaux’s Swift migratory connectivity

4
5

Matthew W. Reudink1*, Steven L. Van Wilgenburg2, Lauren Steele1, Andrew G. Pillar1,

6

Peter P. Marra3 and Ann E. McKellar2

7
8

1

9

Canada.

10

2

Environment Canada, 115 Perimeter Road, Saskatoon, Saskatchewan, Canada.

11

3

Migratory Bird Center, Smithsonian Conservation Biology Institute, Washington, District of

12

Columbia, United States of America

13

* Corresponding author: mreudink@tru.ca

Department of Biological Sciences, Thompson Rivers University, Kamloops, British Columbia,

14
15
16
17
18

1

19

ABSTRACT

20

The strength of migratory connectivity between breeding, stopover, and wintering areas can have

21

important implications for population dynamics, evolutionary processes, and conservation. For

22

example, patterns of migratory connectivity may influence the vulnerability of species and

23

populations to stochastic events. For many migratory songbirds, however, we are only just

24

beginning to understand patterns of migratory connectivity. Here, we investigate the potential

25

strength of migratory connectivity within a population of Vaux’s Swifts (Chaetura vauxii).

26

Vaux’s Swifts, like many aerial insectivores, are currently experiencing population declines, and

27

a mass mortality event at a spring migratory roost on Vancouver Island, British Columbia,

28

Canada, resulted in the death of over 1,000 individuals, representing some 2% of the British

29

Columbia population. From these individuals, we examined variation in three stable-isotopes

30

(δ2H, δ13C, and δ15N) from claw samples in order to determine whether spring migrants showed

31

inherent isotopic similarity of the habitats they used on their Mexican and Central American

32

wintering grounds. Our results indicated the presence of two to three broad isotopic clusters,

33

suggesting that Vaux’s Swifts migrating through Vancouver Island most likely originated from

34

two or three over-wintering locales or habitat types. We found no evidence of sex- or

35

morphology-based segregation, suggesting that these different groups likely share a similar over-

36

wintering ecology and thus may be equally vulnerable to stochastic events or habitat loss on the

37

wintering grounds. Our results highlight the need for more studies on the non-breeding season

38

ecology and migratory connectivity of this species.

39

Keywords: Vaux’s Swift, migratory connectivity, cluster analysis, stable isotope, conservation,

40

roost

2

41

INTRODUCTION

42

Migratory birds move annually between breeding, wintering, and stopover sites that can be

43

separated by hundreds or thousands of kilometers. As such, they face an array of both natural and

44

human-mediated environmental challenges. Thus, for effective conservation, it is essential to

45

take an annual cycle approach that addresses factors that influence populations throughout the

46

year and across broad geographic scales (Webster 2002, Webster and Marra 2005). Stressors

47

occurring during one phase of the annual cycle can have carry-over effects to subsequent phases

48

of the annual cycle, affecting both individual- and population-level dynamics (Marra et al. 1998,

49

Sillett et al. 2000, Reudink et al. 2009a). Population level carry-over effects can manifest through

50

density-dependent population regulation (Fretwell 1972, Ratikainen et al. 2008), large-scale

51

climatic cycles (Sillett et al. 2000, Wilson et al. 2011), and habitat loss (Norris 2005). Examples

52

of carry-over effects influencing individual fitness include wintering habitat-mediated

53

differences in reproductive success (Reudink et al. 2009a) and natal dispersal (Studds et al.

54

2008). Understanding patterns of migratory connectivity, defined as the extent to which

55

individuals from the same wintering site migrate to the same breeding site, and vice versa (Marra

56

et al. 2010), may be important for understanding species- or population-specific responses to

57

anthropogenic disturbances (e.g., land-use change, agricultural development), as well as large-

58

scale selective pressures such as climate change (Webster et al. 2002).

59

Migratory populations may be especially vulnerable to stochastic events when they

60

display high migratory connectivity, particularly if population size is also small. For example,

61

Kirtland’s Warblers (Setophaga kirtlandii) exhibit extremely strong connectivity between

62

breeding populations in Michigan, USA, and wintering populations in the Bahamas and Turks

63

and Caicos (Bocetti et al. 2014). As a consequence, the species experiences pronounced carry-

3

64

over effects whereby drier winters delay arrival and nest initiation on the breeding grounds,

65

ultimately resulting in fewer offspring fledged (Rockwell et al. 2012). Intensive management of

66

the species on the breeding grounds has led to population increases, but continued long-term

67

recovery is likely dependent on active management in the Bahamas as well (Wunderle et al.

68

2010), especially as climate change is predicted to increase drought severity in the Caribbean

69

(Rockwell et al. 2012).

70

For small migratory birds, making geographic connections for individuals and populations

71

between different phases of their annual cycle can be exceedingly difficult. In recent years,

72

information gleaned from traditional bird ringing has been greatly enhanced with the use of

73

intrinsic markers such as genetic (e.g., Ruegg et al. 2014) and stable isotope (Hobson 2011)

74

techniques, as well as combinations of these techniques (Chabot et al. 2012, Rundel et al. 2013).

75

Stable isotopes of hydrogen (δ2H) in particular have been instrumental in migratory connectivity

76

research owing to geographically predictable patterns of isotopic variation that are reflected in

77

animal tissue and thus provide markers of origin for where the tissue was grown (Hobson et al.

78

2012a). Such intrinsic markers have the advantage that birds need only be captured once, and

79

they may be especially useful as a first step to understanding patterns of migratory connectivity

80

in poorly-studied species or in remote areas (Hobson et al. 2014, Pekarsky et al. 2015). The

81

advent of new extrinsic marker technologies such geolocators and GPS loggers has led to higher

82

resolution information on migratory behaviour for certain species (Stutchbury et al. 2009).

83

However, use of these extrinsic markers is complicated by the need to recover data loggers and

84

remains impractical and often prohibitively expensive for large-scale studies (Arlt et al. 2013,

85

Bridge et al. 2013, Hobson et al. 2014). In addition, extrinsic markers cannot be used to infer

86

historic patterns of migratory connectivity. In contrast, intrinsic markers can be extremely useful

4

87

for forensic studies, including dietary reconstruction (Nocera et al. 2012, Blight et al. 2014) and

88

the inference of migratory connectivity from historic specimens (Hobson et al. 2010). Given the

89

paucity of information on population- or region-specific migratory connectivity for many species

90

of conservation concern, there is a pressing need to apply intrinsic and/or extrinsic marker

91

approaches to evaluate factors limiting populations throughout their annual cycle (Hobson et al.

92

2014).

93

Like many other species of aerial insectivore, breeding populations of the Vaux’s Swift

94

(Chaetura vauxii) have undergone significant declines (Nebel et al. 2010). Recent estimates from

95

the North American Breeding Bird Survey suggest a population decline of -2.2%/year in Canada

96

since the 1970s (Environment Canada 2014). Vaux’s Swifts breed throughout western North

97

America, but exhibit an unusual roosting strategy during migratory journeys between their

98

western North American breeding grounds and their southern Mexican and Central American

99

wintering grounds (Bull and Collins 2007). During migration, groups of Vaux’s Swifts roost

100

communally in tree cavities or, more commonly, in abandoned industrial chimneys, where they

101

can remain for several days to weeks (Bull and Collins 2007). Migratory roosts may contain

102

hundreds to thousands of individuals. During winter, Vaux’s Swifts also appear to occupy large

103

roost sites (Bull and Collins 2007). However, little is known about migratory connectivity for

104

Vaux’s Swift and it remains unknown whether swifts from particular wintering areas migrate

105

north together and stop at the same roost sites (strong migratory connectivity), or whether

106

migratory roost sites are comprised of individuals arriving from multiple locations across their

107

winter range (weaker migratory connectivity). The degree of connectivity can have important

108

implications for conservation (e.g. Sheehy et al. 2011), and if large concentrations of individuals

109

winter together and move together throughout migration, they may be particularly vulnerable to

5

110

stochastic events. In this study, we used intrinsic markers to study putative migratory

111

connectivity in migrating Vaux’s Swifts. We sampled claws (grown during the over-wintering

112

period) from a large group of communally-roosting Vaux’s Swifts that perished due to accidental

113

causes at a spring migratory roost site on Vancouver Island, British Columbia, Canada. We used

114

cluster analyses based on multiple stable isotope (δ2H, δ13C, and δ15N) markers to examine

115

support for the existence of a single versus multiple wintering origins in this population of swifts.

116
117

METHODS

118

Field Methods

119

On May 9, 2012, a mass mortality event occurred at a migratory roost in Cumberland, British

120

Columbia, Canada, used annually by Vaux’s Swifts during spring migration. Approximately

121

1,350 Vaux’s Swifts perished from suffocation after being trapped in the roost. That single

122

mortality event represented a loss of between 1.5-2.7 % of published population size estimates

123

(Summers and Gebauer 1995, Partners in Flight Science Committee 2013) for Vaux’s Swifts in

124

British Columbia. From those individuals, we randomly sub-sampled 98 individuals for analysis.

125

From each individual we measured wing length, tail length, and tail pin length, and we sexed

126

each individual via laparoscopy. We then sampled 2mm (0.30 to 0.40 mg) of tissue from the tip

127

of the central claw, ensuring we avoided the quick, of each foot from each individual for isotope

128

analysis. Though claws grow continuously and, due to their conical growth pattern, incorporate

129

some material grown at slightly different times, analysis of the claw tip should accurately reflect

130

the isotopic environment over an extended time period (weeks to months) preceding sampling

131

(Mazerolle and Hobson 2005, Hahn et al. 2012). For instance, based on changes in δ2H after

132

arrival on the breeding grounds, Fraser et al. (2008) estimated isotopic values in claws of

6

133

Parulids may reflect the non-breeding area for 3-7 weeks after the birds arrive on the breeding

134

grounds. Hahn et al. (2014) also provide empirical evidence demonstrating that in Palearctic-

135

African migratory passerines the distal claw tip should reflect isotopic environments over a few

136

months prior to sampling, with typical claw growth rates of 0.03-0.05 mm d-1. Though data on

137

migration timing of Vaux’s Swifts are sparse, major flight passages occur during mid-April to

138

late May in California (Small 1994), with birds arriving in Oregon from late April to late May

139

(Bull and Collins 2007 and references therein). Thus, claws sampled from birds that perished

140

during spring migration on May 9 should reflect wintering conditions, as has been demonstrated

141

with passerines captured both on migration (Bearhop et al. 2004) and upon arrival on the

142

breeding grounds (Reudink et al. 2009a,b).

143
144

Stable Isotope Analysis

145

We analyzed the stable isotope ratios of three naturally-occurring elements that are incorporated

146

predictably into an animal’s diet. First, we analyzed δ2H which has been used extensively in

147

studies of avian migration and connectivity because it is linked to both latitude and elevation,

148

with 2H being relatively less abundant at more northern latitudes and higher elevations (Hobson

149

2011). Second, we used δ13C, which varies in animal tissues with habitat, where individuals

150

occupying habitats with a higher proportion of C4 plants or under higher water stress exhibit less

151

negative δ13C signatures (Lajtha and Marshall 1994, Cerling et al. 1997; Still and Powell 2010).

152

Finally, we used δ15N, where 15N is preferentially incorporated into the tissues of consumers and

153

thus biomagnifies with increasing trophic levels, leading to an increase in δ15N signatures (Post

154

2002, Poupin et al. 2011). In addition, δ15N may reflect the relative aridity of a biome, with δ15N

155

being negatively correlated with rainfall and positively correlated with temperature (Sealy et al.

156

1987, Craine et al. 2009). All stable isotope analyses were conducted at the Smithsonian
7

157

Institution OUSS/MCI Stable Isotope Mass Spectrometry Facility in Suitland, Maryland, USA.

158

Claws were washed in a 2:1 chloroform:methanol solution, then air-dried and allowed to

159

acclimate to lab atmospheric conditions in a fume hood for 72 hours prior to sample preparation.

160

Samples were pyrolyzed in a Thermo TC/EA elemental analyzer (Thermo Scientific, Waltham,

161

Massachusetts, USA) at 1,350 °C and analyzed using a Thermo Delta V Advantage isotope ratio

162

mass spectrometer. For stable hydrogen isotope analysis, we ran four calibrated standards for

163

every 10 samples, including the hydrogen standard International Atomic Energy Agency CH-7

164

and three additional standards (KHS: -54.1 ± 2.3‰, CBS: -197.3 ± 2.0‰, Spectrum keratin: -

165

121.6 ± 2.5‰). Non-exchangeable δ2H values were corrected to keratin standards following

166

Wassenaar and Hobson (2003), and were repeatable to within 3 ± 2‰ (mean ± SD, n = 10). For

167

stable carbon and nitrogen analysis, we ran two in-house standards (acetanalide and urea) for

168

every 10 samples. Stable isotope values are expressed in parts per thousand (‰) deviations from

169

international standards VSMOW (hydrogen), PDB (carbon), and air (nitrogen) by the following

170

equation:

171

dX = {[(Runknown-Rstandard)-1] x 1000},

172

where X is the isotope ratio of interest (δ2H, δ13C, or δ15N) and R is the corresponding ratio

173

(2H:1H, 13C:12C or 15N:14N). Carbon and Nitrogen samples were repeatable to within ± 0.2 ‰

174

based on repeated measurements of standards.

175
176

Statistical Analysis

177

There are many different methods available for determining the optimal number of clusters in a

178

dataset. For this reason, we made use of the package ‘NbClust’ in program R (R Development

179

Core Team, Vienna, Austria), which provides the user with results from 30 indices aimed at

8

180

determining the number of clusters into which the data should be split (Charrad et al. 2014).

181

Clustering validity indices combine information such as intra and inter-cluster variation,

182

geometric or statistical properties of the data, and dissimilarity or similarity measurements. For a

183

detailed description of all 30 indices, see Charrad et al. (2014). Two widely-used clustering

184

algorithms are available in the NbClust package: k-means and hierarchical agglomerative

185

clustering. In k-means clustering, observations are assigned to initial cluster centers, which are

186

then iteratively updated until the cluster centers no longer change and the within-cluster sum of

187

squares is minimized (MacQueen 1967). In hierarchical agglomerative clustering, each

188

observation begins in its own cluster, and pairs of clusters are joined based on a distance measure

189

and an agglomeration criterion (Székely and Rizzo 2005). Because of the differences in these

190

clustering approaches and a lack of a priori reasons to hypothesize a set number of clusters in

191

our data, we used both clustering algorithms. We used a Euclidean distance (square distance

192

between two vectors; Seber 1984) in both clustering algorithms and examined the relative

193

support for division of the data into 2-10 clusters based on the three isotopic markers (δ2H, δ13C,

194

and δ15N). For hierarchical agglomerative clustering, we used the Ward agglomeration method

195

(Ward 1963), which minimizes the total within-cluster variance. We also repeated the above

196

procedures after removing six multivariate outliers, assessed using the robust Mahalanobis

197

distance (Varmuza and Filzmoser 2009).

198

After the optimal number of clusters was identified by each clustering approach, we

199

examined the partitioning of observations (individual birds) into clusters. We performed

200

multivariate analysis of variance (MANOVA) to verify significant differences among the

201

selected clusters in δ2H, δ13C, and δ 15N ratios. We then used a Pearson’s chi square test to

202

examine whether the proportion of claws from male and female birds differed among clusters,

9

203

and we used a t test or analysis of variance (ANOVA) to examine whether wing length or tail

204

length varied among clusters.

205

Finally, we used a model-based hierarchical clustering procedure within the package

206

‘mclust’ in R (Fraley et al. 2012) to determine whether the optimal number(s) of clusters selected

207

by the above procedures was preferred over no clustering at all. Mclust uses an expectation-

208

maximization (EM) algorithm to estimate the finite mixture models that correspond to different

209

numbers of clusters, and uses the Bayesian information criterion (BIC) to select the best number

210

of clusters. Importantly, a single cluster can also be considered. We evaluated the support for 1-

211

10 clusters.

212

The δ2H, δ13C, and δ15N isotope ratios were standardized prior to analysis by subtracting by

213

their mean and dividing by their standard deviation. Standardization is recommended in cluster

214

analysis when variables are on different scales to minimize the effects outliers and so cluster

215

formation is not overly influenced by variables with greater absolute variation (Milligan and

216

Cooper 1988). Although isotope ratios represent the ratio of heavy to light isotopes in a sample

217

divided by the ratio of a standard, and are always expressed in the same units (‰), differences in

218

the natural abundances and fractionation of heavy and light isotopes among elements lead to

219

differences in the magnitude of the range of typical δ values (Fry 2006). Thus, standardization

220

was useful in giving equal weight to variation among the three isotope ratios. All analyses were

221

performed in R version 3.1.1 and means are shown as mean ± SD.

222
223

RESULTS

224

Stable isotopic values (δ2H, δ13C, and δ15N) were available from claws of 98 individual swifts.

225

All three elements showed a considerable variation in all three isotopes (Supplementary Material

10

226

Table S1). Mean isotope ratios were -37.9 ± 12.4 ‰ for δ2H, -21.6 ± 0.6 ‰ for δ13C, and 7.3 ±

227

0.9 ‰ for d15N. Correlations among isotope ratios were r = -0.09 (p = 0.37) for δ2H and δ13C, r =

228

-0.14 (p = 0.17) for δ2H and δ15N, and r = 0.53 (p < 0.001) for δ13C and δ15N. Of those 98

229

individuals with isotope data, we were unable to determine sex for one individual. There was a

230

nearly-even number of males and females in our sample (n = 47 females and 50 males). Males

231

and females did not differ in δ2H (t95 = -0.02, p = 0.98), δ13C (t95 = -0.98, p = 0.33), or δ15N (t95 =

232

-0.18, p = 0.86). Mean wing length was 113.5 ± 2.8 mm and did not differ between males and

233

females (males: 113.4 ± 2.7, females: 113.7 ± 2.8; t95 = -0.46, p = 0.65). Wing length was not

234

correlated with δ2H (r = -0.06, p = 0.59), d13C (r = 0.16, p = 0.12), or δ15N (r = 0.16, p = 0.12).

235

Mean tail length was 36.2 ± 2.3 mm and did not differ between males and females (males: 35.8 ±

236

1.8, females: 36.5 ± 2.7; t95 = -1.53, p = 0.13). Tail length was not correlated with δ2H (r = 0.00,

237

p = 0.99) or d13C (r = -0.07, p = 0.49), but it was negatively correlated with δ15N (r = -0.23, p = -

238

0.02).

239

When using the full dataset, 11 out of 30 clustering validity indices (37%) proposed two as

240

the optimal number of clusters in the k-means clustering procedure. The next highest-ranked

241

numbers of clusters were three and four, which received support from six (20%) and five (17%)

242

clustering indices, respectively. Three was deemed the optimal number of clusters by the

243

majority (10 or 33%) of indices in the hierarchical agglomerative procedure, but a large

244

proportion of indices (9 or 30%) also supported two clusters. The only alternative clustering that

245

received support was a 10 cluster solution with support from five (17%) indices. Results of the k-

246

means clustering procedure were similar when outliers were removed, with two deemed the

247

optimal number of clusters by the majority (9 or 30%) of indices. For the hierarchical

248

agglomerative procedure, two was identified as the optimal number (8 or 27% of indices), with

11

249

three receiving the second-most support (5 or 17% of indices). Overall, these results suggest that

250

the isotope data grouped most naturally into two or three clusters, and thus we used the full

251

dataset for the remainder of the analyses. Although the selection of the optimal number of

252

clusters was not unanimous among indices, similar results are obtained from simulated datasets

253

even when distinct non-overlapping clusters are used (Milligan & Cooper 1985, Charrad et al.

254

2014). Furthermore, according to the BIC, the model-based clustering procedure suggested that

255

the optimal number of clusters was two (Supplementary Material Table S2).

256

We examined the classification of individual claws into two or three clusters using the

257

results from k-means clustering or hierarchical agglomerative clustering, respectively. The

258

clusters were significantly different from one another for all three isotope ratios simultaneously

259

when partitioned into two (MANOVA F3,93 = 52.5, p < 0.001) or three clusters (F3,93 = 36.4, p <

260

0.001). The two cluster solution resulted from a split between swifts with higher δ15N and δ13C

261

ratios and those with lower δ15N and δ13C ratios, with little influence of d2H on clustering (Table

262

1). The three cluster solution suggested an additional division of the swifts with low δ15N and

263

δ13C ratios into those with higher or lower δ2H ratio (Figures 1, 2). There was no association

264

between sex and cluster membership when the claws were grouped into two (χ21 = 0, p = 1) or

265

three (χ22 = 1.70, p = 0.42) clusters. There was no association between wing length and cluster

266

membership for two (t96 = 0.62, p = 0.54) or three (ANOVA F2,95 = 0.70, p = 0.50) clusters.

267

Similarly, there was no association between tail length and cluster membership for two (t96 = -

268

0.48, p = 0.64) or three (ANOVA F2,95 = 1.60, p = 0.21) clusters.

269

12

270

DISCUSSION

271

Patterns of migratory connectivity in long-distance migratory birds can have important

272

implications for population dynamics, evolutionary processes, and effective conservation

273

strategies (Webster and Marra 2005). Using multiple stable isotopes (δ13C, δ15N, and δ2H) from

274

claw samples of Vaux’s Swifts at a spring migratory roost, our goal was not to assign individuals

275

to a particular over-wintering locality or localities, which would have been exceedingly difficult

276

due to poor resolution in isoscapes of Mexico and Central America (Bowen et al. 2005) and our

277

lack of known-origin tissues from these areas. Rather we asked whether Vaux’s Swifts at a single

278

migratory roost likely came from a single winter location, which would indicate stronger

279

migratory connectivity, or multiple winter locations, which would indicate weaker migratory

280

connectivity. We found evidence in support of two or possibly three broad isotopic clusters on

281

the over-wintering grounds from the single migratory roost on Vancouver Island, British

282

Columbia, Canada. While alternative explanations for our observations are possible (see below),

283

we tentatively interpret our findings as likely demonstrating differences in over-wintering

284

locations used by Vaux’s Swifts in our sample.

285

Understanding patterns of migratory connectivity in small, long-distance migratory birds

286

has been exceedingly challenging, but the use of stable isotope analysis over the past two

287

decades has revolutionized our ability to establish patterns of connectivity (Hobson 2005).

288

Generally, this approach relies on the assignment of tissues to spatially explicit isoscapes

289

generated from precipitation isoscapes (Bowen and West 2008) or known-origin individuals

290

sampled across their range (e.g., Hobson et al. 2009c, Hobson et al. 2012b). Recently, these

291

methods have been enhanced through the use of Bayesian statistical methods and GIS-based

292

models of precipitation isoscapes, as well as additional information such as band recoveries, to

13

293

improve assignments by creating probabilistic regions of origin (Hobson et al. 2009a, b, Van

294

Wilgenburg and Hobson 2011). However, most studies focus on assigning breeding origins to

295

winter-captured individuals, rather than vice versa, because breeding ground isoscapes,

296

particularly in North America, are better delineated than winter ground isoscapes (Bowen et al.

297

2005). This poses a problem for assigning winter origins to individuals sampled within breeding

298

or migratory populations, such as in our study. Furthermore, the particular tissue chosen for

299

sampling must be reflective of the area of origin that is of interest to the researcher. This again

300

poses difficulties for assigning winter origins when many North American bird species,

301

including Vaux’s Swifts, moult their feathers on or near the breeding grounds (Pyle 1997; but see

302

for e.g. Kelly et al. 2008, Quinlan and Green 2011). Thus, in our study, we chose to sample

303

claws because they reflect conditions several weeks prior to sampling, a time during which the

304

birds should have been present on the wintering grounds (Bearhop 2004, Reudink et al. 2009a,b).

305

Despite the relatively high variation in all three isotopic signatures (δ2H range: 66 ‰;

306

δ15N range: 5.8 ‰; δ13C range: 3.5 ‰; Supplementary Material Table S1), our analysis provided

307

support for Vaux’s Swift claws fitting into two or three isotopically distinct clusters (Figure 2),

308

with strongest support for two clusters when outliers were removed. When partitioned into two

309

clusters, the claws separated based on those with higher δ13C and δ15N and those with lower d13C

310

and d15N. Variation in δ13C and δ15N most likely relate to broad-scale habitat differences, as

311

food-web δ13C signatures are strongly related to the distribution of C3 and C4 vegetation (Lajtha

312

and Marshall 1994), and food-web δ15N signatures are associated with soil exposure and climate

313

(Nadelhoffer and Fry 1994). While we cannot exclude the possibility that birds were separated in

314

part on the basis of dietary shifts and niche specialization, broad-scale differences in δ13C, δ15N,

315

and δ2H appear to be consistent with Vaux’s Swift using habitats that follow climatic/moisture

14

316

gradients. Regardless of the clustering algorithm, birds associated with cluster 2 were enriched in

317

13

318

were also enriched in 15N by ~ 1.6 ‰ compared to birds from cluster 1 (Table 1). Assuming δ13C

319

follows similar gradients to those reported elsewhere (e.g. Marra et al. 1998), these data would

320

appear to suggest that birds in cluster 2 originated from hotter and drier habitats than birds in

321

cluster 1.

C by ~0.9 ‰ over individuals from cluster 1. Simultaneously, birds associated with cluster 2

322

The agglomerative clustering approach suggested that birds could be further partitioned

323

into a third cluster on the basis of splitting claws that had both low d13C and d15N values based

324

on those with higher δ2H (cluster 3) and versus lower δ2H values (cluster 1, see Supplementary

325

Material Table S1). Latitudinal and altitudinal gradients in δ2H are well-established and among

326

the strongest of isotopic gradients (Hobson 2005), and our results thus suggest that individuals

327

that had low δ13C and δ15N were likely further segregated along an latitudinal or possibly an

328

altitudinal gradient. Although a species-specific δ2H claw tissue isoscape is lacking for the

329

Vaux’s Swift over-wintering range, a δ2H feather isoscape developed for House Sparrows

330

(Passer domesticus) in Mexico showed more positive δ2H signatures in the northeast of the

331

country and more negative signatures in the south and high altitude regions (Hobson et al.

332

2009c). Overall, our results indicate that individuals from the migratory population of swifts on

333

Vancouver Island may have originated from a small number of different latitudes and/or habitats.

334

Breeding populations that are more strongly connected to specific wintering areas,

335

particularly those experiencing degradation, may be most vulnerable to population declines

336

(Dolman and Sutherland 1995, Webster and Marra 2005), although a recent study found that the

337

strength of connectivity in breeding populations of Barn Swallows (Hirundo rustica) in North

338

America was unrelated to their population trend (García-Pérez and Hobson 2014). Our data

15

339

suggest that further research to estimate the strength of connectivity across breeding populations

340

of Vaux’s Swifts is now warranted, and might help explain patterns of decline if extrinsic

341

markers could be deployed along a gradient of declining versus increasing populations (sensu

342

Fraser et al. 2012 and Hobson et al. 2015).

343

We also asked whether swifts showed evidence of segregation based on sex or

344

morphology during the over-winter period, as has been observed in several other passerines, such

345

as Hooded Warblers (Setophaga citrine; Morton 1990), Prairie Warblers (Setophaga discolor;

346

Latta and Faaborg 2001), and American redstarts, that segregate based on habitat, and White-

347

throated Sparrows (Zonotrichia albicollis; Mazerolle and Hobson 2007) and Hermit Thrushes

348

(Catharus guttatus; Stouffer and Dwyer 2003), that segregate based on latitude. For example,

349

American Redstarts (Setophaga ruticilla) exhibit marked age- and sex-biased habitat

350

segregation, with older males disproportionately inhabiting high-quality mangrove forests

351

(Parish and Sherry 1994, Marra 2000) – a characteristic also easily discerned via stable isotope

352

analysis of muscle (Marra et al. 1998), blood (Norris et al. 2004), and claw samples (Reudink et

353

al. 2009a,b). Furthermore, larger females were more likely to inhabit high-quality mangrove

354

territories due to dominance-mediated habitat segregation (Marra 2000). Latitudinal segregation,

355

on the other hand, may be related to trade-offs between body size or wing size/shape and

356

migration distance (García Peiró 2003; Mazerolle and Hobson 2007). Our results suggest no

357

evidence for sex- or morphology-based segregation, at least when considering wing and tail

358

length, and indeed there was no evidence of differences in wing size between males and females,

359

suggesting that male and female Vaux’s Swifts likely share a similar over-wintering ecology and

360

thus may be equally vulnerable to stochastic events or habitat loss.

16

361

The utility of triple-isotope isoscapes for assigning origins to South American and

362

African wintering birds has recently been demonstrated (Hobson et al. 2012b, García-Pérez and

363

Hobson 2014), and our results indicate that the approach has potential for identifying Mexican

364

and Central-American wintering clusters as well. Taken together, our results highlight the need

365

for additional research on the over-wintering ecology and behaviour of this poorly-studied long-

366

distance migrant and the need for calibrating tissue specific isoscapes to better assign Nearctic-

367

Neotropical migratory birds to wintering localities. Our data suggest that birds migrating through

368

Vancouver Island likely over-winter in two to three isotopically similar regions or habitat types,

369

and these predictions should now be validated. Wintering ground sampling would facilitate

370

nominal assignment (sensu Wunder 2012) to two or three spatially delineated geographic

371

regions, or alternatively might facilitate spatially explicit multivariate assignment approaches

372

(e.g., García-Pérez and Hobson 2014, Veen et al. 2014).

373

Recent work highlights the potential for keratinous tissues to vary isotopically with diet

374

(Fraser et al. 2011, Voigt et al. 2013, Voigt et al. in press, Soto et al. 2013) and microhabitat

375

(Fraser et al. 2011). Thus, while we interpret our results primarily in the context of migratory

376

connectivity, we cannot preclude the possibility that multivariate differences in the isotopic

377

composition of claws could stem from variation in microhabitat, behaviour, or dietary

378

preferences among individuals. Given the high motility of Vaux’s swift while foraging on the

379

wing (Bull et al. 2007), it is plausible that our results could also stem from variation in degree of

380

foraging over aquatic versus terrestrial or forested versus grassland or agricultural habitats. Like

381

other aerial insectivores, Vaux’s swifts forage broadly over forest canopy, meadows and open

382

water (Bull and Collins 2007) and their diets likely reflect the composition of aerial plankton

383

from major insect emergences from both the aquatic and terrestrial environments. Within their

17

384

wintering grounds, these major habitat types should be relatively isotopically distinct owing to

385

major isotopic differences in δ13C and δ15N between forest canopy (primarily using a C3

386

phytosynthetic pathway) versus grasslands and pastures dominated by C4 plants (Still and Powell

387

2010). It is important to note that the distribution of forest versus grassland, pasture and

388

agricultural lands displays a great deal of geographic structure (Dixon et al. 2014, Ellis and

389

Ramankutty 2008), and thus it is probable that isotopic differences due to differences in habitat

390

selection would also covary with any spatial segregation. Regardless of whether the patterns we

391

observed owe solely to differences in migratory connectivity or variation in local habitat

392

selection, our results point to relatively limited variation in over-wintering niches of Vaux’s

393

swifts. Thus, a better understanding of the overwintering ecology Vaux’s swifts is necessary to

394

inform conservation. Wintering ground work on habitat use and diets of Vaux’s swifts or

395

performing multi-isotope assays on claws from birds fitted with miniaturized GPS tags

396

(Hallworth and Marra 2015) would shed light on the mechanistic explanations for our

397

observations.

398

Like most migratory aerial insectivores (Nebel et al. 2010), Vaux’s Swifts are in decline.

399

Given the unique roosting ecology of Vaux’s Swifts, they are particularly susceptible to

400

stochastic events as exemplified by this one mass mortality, which may represent a loss of ~1.5 –

401

2.7% of the British Columbia population of Vaux’s Swifts in a single evening. This exceptional

402

vulnerability coupled with increasing habitat loss and alteration in the tropics (Hansen et al.

403

2013) make the need for further research to determine patterns of migratory connectivity and

404

identify critical winter and migratory roost sites crucial for future conservation planning.

405
406

18

407

Acknowledgements We thank S. Joly for assistance in data collection. Funding was provided

408

by a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant

409

and a TRU Internal Research Fund Award to M.W.R. and a National Science Foundation (NSF)

410

Grant to P.P.M. Birds were collected under a Canadian salvage permit (BC-SA-0049-12).

411

19

412

LITERATURE CITED

413

Arlt, D., M. Low, and T. Part (2013). Effect of Geolocators on Migration and Subsequent

414

Breeding Performance of a Long-Distance Passerine Migrant. PLOS ONE 8:e82316.

415

Bearhop, S., G. M. Hilton, S. C. Votier, and S. Waldron (2004). Stable isotope ratios indicate

416

that body condition in migrating passerines is influenced by winter habitat. Proceedings

417

of the Royal Society B: Biological Sciences 271:S215–S218.

418

Blight, L. K., K. A. Hobson, T. Kurt Kyser, and P. Arcese (2014). Changing gull diet in a

419

changing world: a 150-year stable isotope (d13C, d15N) record from feathers collected in

420

the Pacific Northwest of North America. Global Change Biology 21:1497-1507.

421

Bocetti, C. I., M.D. Deahn and H. F. Mayfield (2014). Kirtland's Warbler (Setophaga kirtlandii),

422

The Birds of North America Online (A. Poole, Ed.). Ithaca: Cornell Lab of Ornithology;

423

Retrieved from the Birds of North America Online:

424

http://bna.birds.cornell.edu/bna/species/019

425
426
427

Bowen, G. J., L. I. Wassenaar, and K. A. Hobson (2005). Global application of stable hydrogen
and oxygen isotopes to wildlife forensics. Oecologia 143:337-348.
Bowen, G.J., and J.B. West (2008). Isotope landscapes for terrestrial migration research. In:

428

Tracking animal migration with stable isotopes. Eds: K.A. Hobson and L.I. Wassenaar.

429

Academic Press, Amsterdam.

430

Bridge, E. S., J. F. Kelly, A. Contina, R. M. Gabrielson, R. B. MacCurdy, and D. W. Winkler

431

(2013). Advances in tracking small migratory birds: a technical review of light-level

432

geolocation. Journal of Field Ornithology 84:121-137.

20

433

Bull, E. L., and C. T. Collins (2007). Vaux’s Swift (Chaetura vauxi), The Birds of North

434

America Online (A. Poole, Ed.). Ithaca: Cornell Lab of Ornithology; Retrieved from the

435

Birds of North America Online: http://bna.birds.cornell.edu/bna/species/077

436

Cerling, T. E., J. M. Harris, B. J. MacFadden, M. G. Leaky, J. Quade, V. Eisenmann, and J. R.

437

Ehrlinger (1997). Global vegetation change through the Miocene/Pliocene boundary.

438

Nature 389:153-158.

439

Chabot, A. A., K. A. Hobson, S. L. Van Wilgenburg, G. J. McQuat, and S. C. Lougheed (2012).

440

Advances in Linking Wintering Migrant Birds to Their Breeding-Ground Origins Using

441

Combined Analyses of Genetic and Stable Isotope Markers. PLOS ONE 7:e43627.

442

Charrad, M., N. Ghazzali, V. Boiteau, and A. Niknafs (2014). NbClust: an R package for

443

determining the relevant number of clusters in a data set. Journal of Statistical Software

444

61:1-36.

445

Craine, J. M., A. J. Elmore, M. P. M. Aidar, M. Bustamante, T. E. Dawson, E. A. Hobbie, A.

446

Kahmen, M. C. Mack, K. K. McLauchlan, A. Michelsen, G. B. Nardoto, L. H. Pardo, J.

447

Peñuelas, P. B. Reich, E. A. G. Schuur, W. D. Stock, P. H. Templer, R. A. Virginia, J. M.

448

Welker, and I. J. Wright (2009). Global patterns of foliar nitrogen isotopes and their

449

relationships with climate, mycorrhizal fungi, foliar nutrient concentrations, and nitrogen

450

availability. New Phytologist 183:980–992.

451

Dixon, A. P., D. Faber-Langendoen, C. Josse, J. Morrison, and C. J. Loucks (2014). Distribution

452

mapping of world grassland types. Journal of Biogeography 41: 2003-2019.

453

Dolman, P. M., and W. J. Sutherland (1995). The response of bird populations to habitat loss.

454

Ibis 137:S38–S46.

21

455
456
457
458

Ellis, E. C., and N. Ramankutty (2008). Putting people in the map: anthropogenic biomes of the
world. Frontiers in Ecology and the Environment 6: 439–447.
Environment Canada (2014). North American Breeding Bird Survey - Canadian Trends Website,
Data-version 2012. Environment Canada, Gatineau, Quebec.

459

Fraley, C., A. E. Raftery, T. B. Murphy, and L. Scrucca (2012). mclust Version 4 for R: Normal

460

Mixture Modeling for Model-Based Clustering, Classification, and Density Estimation.

461

Technical Report No. 597, Department of Statistics, University of Washington.

462

Fraser, K. C., T. K. Kyser, R. J. Robertson, and L. M. Ratcliffe (2008). Season patterns in

463

hydrogen isotopes of claws from breeding wood-warblers (Parulidae): utility for

464

estimating migratory origins. Avian Conservation and Ecology 3:2.

465

Fraser, K. C., E. A. McKinnon, A. W. Diamond, and L. Chavarría (2011). The influence of

466

microhabitat, moisture and diet on stable-hydrogen isotope variation in a Neotropical

467

avian food web. Journal of Tropical Ecology 27: 563-572.

468

Fraser, K. C., B. J. Stutchbury, C. Silverio, P. M. Kramer, J. Barrow, D. Newstead, N. Mickle, B.

469

F. Cousens, J. C. Lee, D. M. Morrison, T. Shaheen, P. Mammenga, K. Applegate, and J.

470

Tautin (2012). Continent-wide tracking to determine migratory connectivity and tropical

471

habitat associations of a declining aerial insectivore. Proceedings of the Royal Society of

472

London B: Biological Sciences, rspb20122207.

473
474
475

Fretwell, S. D. (1972). Populations in a seasonal environment. Princeton, New Jersey: Princeton
University Press.
Fry, B. (2006). Stable Isotope Ecology. Springer, New York, NY.

22

476

García Peiró, I. (2003). Intraspecific variation in the wing shape of the long-distance migrant

477

Reed Warbler Acrocephalus scirpaceus: effects of age and distance of migration. Ardeola

478

50:31-37.

479

García-Pérez, B., and K. A. Hobson (2014). A multi-isotope (δ2H, δ13C, δ15N) approach to

480

establishing migratory connectivity of Barn Swallow (Hirundo rustica). Ecosphere

481

5:art21

482

Hahn, S., B. J. Hoye, H. Korthals, and M. Klaasen (2012). From food to offspring down: tissue-

483

specific discrimination and turn-over of stable isotopes in herbivorous waterbirds and

484

other avian foraging guilds. PLOS One 7:e30242.

485

Hallworth, M. T., and P. P. Marra. 2015. Miniaturized GPS tags identify non-breeding territories

486

of a small breeding migratory songbird. Scientific Reports 5: art 11069.

487

doi:10.1038/srep11069

488

Hahn, S., D. Dimitrov, S. Rehse, E. Yohannes, and L. Jenni (2014). Avian claw morphometry

489

and growth determine the temporal pattern of archived stable isotopes. Journal of Avian

490

Biology 45:202-207.

491

Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D.

492

Thau, S. V. Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini,

493

C. O. Justice, and J. R. G. Townshend (2013). High-resolution global maps of 21st-

494

century forest cover change. Science 342:850-853.

495
496
497

Hobson, K. A. (2005). Stable isotopes and the determination of avian migratory connectivity and
seasonal interactions. Auk 122:1037-1048.
Hobson, K. A. (2011). Isotopic ornithology: a perspective. J. Ornithol. 152 (Suppl 1):S49–S66.

23

498

Hobson, K. A., M. B. Wunder, S. L. Van Wilgenburg, R. G. Clark, & L. I. Wassenaar (2009a). A

499

method for investigating population declines of migratory birds using stable isotopes:

500

origins of harvested lesser scaup in North America. PLOS ONE e:7915.

501

Hobson, K. A., H. Lormée, S. L. Van Wilgenburg, L. I. Wassenaar, J. M. Boutin (2009b). Stable

502

isotopes (δD) delineate the origins and migratory connectivity of harvested animals: the

503

case of European woodpigeons. Journal of Applied Ecology 46:572-581.

504

Hobson, K. A., S. L. Van Wilgenburg, K. Larson, & L. I. Wassenaar (2009c). A feather

505

hydrogen isoscape for Mexico. Journal of Geochemical Exploration 102:167-174.

506

Hobson, K. A., R. Greenberg, S. L. Van Wilgenburg, and C. Mettke-Hoffman (2010). Migratory

507

connectivity in the Rusty Blackbird: isotopic evidence from feathers of historical and

508

contemporary specimens. Condor 112:778-788.

509

Hobson, K. A., S. L. Van Wilgenburg, L. I. Wassenaar, K. Larson (2012a). Linking hydrogen

510

(δ2H) isotopes in feathers and precipitation: sources of variance and consequences for

511

assignment to isoscapes. PLOS ONE 7:e35137.

512

Hobson, K. A., S. L. Van Wilgenburg, L. I. Wassenaar, R. L. Powell, C. J. Still, and J. M. Craine

513

(2012b). A multi-isotope (δ13C, δ15N, δ2H) feather isoscape to assign Afrotropical

514

migrant birds to origins. Ecosphere 3:art44

515

Hobson, K. A., S. L. Van Wilgenburg, J. D. Toms, J. Faaborg, C. Rengifo, Y. Aubry, and R. B.

516

Aguilar (2014). Connecting breeding and wintering grounds of Neotropical migrant

517

songbirds using stable hydrogen isotopes: A call for an isotopic atlas of migratory

518

connectivity. Journal of Field Ornithology 85:237-257.

24

519

Hobson, K. A., K. J. Kardynal, S. L. Van Wilgenburg, G. Albrecht, A. Salvadori, M. D. Cadman,

520

F. Liechti, and J. D. Fox. (2015). A continent-wide migratory divide in North American

521

breeding Barn Swallows (Hirundo rustica). PLOS One 10: e0129340.

522

Kelly, J. F., M. J. Johnson, S. Langridge, and M. Whitfield (2008). Efficacy of stable isotope

523

ratios in assigning endangered migrants to breeding and wintering sites. Ecological

524

Applications 18:568-576.

525
526

Latta, S. C. and J. Faaborg (2001). Winter site fidelity of Prairie Warblers in the Dominican
Republic. Condor 103:455-468.

527

Lajtha, K. & Marshall, J. D. (1994). Sources of variation in isotopic composition in plants.

528

In Stable isotopes in ecological and environmental sciences (eds K. Lathja & J. D.

529

Marshall), pp. 1–21. London, UK: Oxford University Press.

530

MacQueen, J. B. (1967). Some methods for classication and analysis of multivariate

531

observations. In LML Cam, J Neyman (eds.), Proceedings of the Fifth Berkeley

532

Symposium on Mathematical Statistics and Probability 1:281-297.

533
534

Marra, P. P., K. A. Hobson, and R. T. Holmes (1998). Linking winter and summer events in a
migratory bird by using stable-carbon isotopes. Science 282:1884–1886.

535

Marra, P. (2000). The role of behavioral dominance in structuring patterns of habitat occupancy

536

in a migrant bird during the nonbreeding season. Behavioral Ecology 11:299–308.

537

Marra, P. P., C. E. Studds, and M. Webster (2010) Migratory connectivity. In: Breed MD, Moore

538

J (eds) Encyclopedia of animal behavior. Academic, Oxford, pp 455–461.

539

Mazerolle, D. F., and K. A. Hobson (2005). Estimating origins of short-distance migrant

540

songbirds in North America: contrasting inferences from hydrogen isotope measurements

541

of feathers, claws, and blood. Condor 107:280–288.

25

542

Mazerolle, D. F., and K. A. Hobson (2007). Patterns of differential migration in white-throated

543

sparrows evaluated with isotopic measurements of feathers. Canadian Journal of Zoology

544

85:413-420.

545
546
547
548

Milligan, G. W., and M. C. Cooper (1985). An examination of procedures for determining the
number of clusters in a data set. Psychometrika 50:159-179.
Milligan, G. W., and M. C. Cooper (1988). A study of standardization of variables in cluster
analysis. Journal of Classification 5:181-204.

549

Morton, E.S. (1990). The evolution of habitat segregation by sex in the Hooded Warbler:

550

experiments on proximate causation. American Naturalist 135: 319-333.

551

Nadelhoffer, K. J., and B. Fry (1994). Nitrogen isotope studies in forest ecosystems. Pages 22–

552

44 in Stable Isotopes in Ecology and Environmental Science (K. Lajtha and R. H.

553

Michener, Eds.). Blackwell Scientific Publications, Oxford.

554

Nebel, S., A. Mills, J. D. McCracken, and P. D. Taylor (2010). Declines of Aerial Insectivores in

555

North America Follow a Geographic Gradient. Avian Conservation and Ecology 5:1.

556

Nocera, J.J., J.M. Blais, D.V. Beresford, L.K. Finity, C. Grooms, L.E., Kimpe, K. Kyser, N.

557

Michelutti, M.W. Reudink, and J.P. Smol. 2012. Historical pesticide applications

558

coincided with an altered diet of aerially foraging insectivorous chimney

559

swifts. Proceedings of the Royal Society B: Biological Sciences 278:3114-3120.

560
561

Norris, D. R. (2005). Carry-over effects and habitat quality in migratory populations. Oikos,
109:178-186.

562

Norris, D. R., P. P. Marra, T. K. Kyser, T. W. Sherry, and L. M. Ratcliffe (2004). Tropical winter

563

habitat limits reproductive success on the temperate breeding grounds in a migratory bird.

564

Proceedings of the Royal Society B: Biological Sciences 271:59–64.

26

565
566

Parrish, J., and T. Sherry (1994). Sexual Habitat Segregation by American Redstarts Wintering
in Jamaica - Importance of Resource Seasonality. Auk 111:38–49.

567

Partners in Flight Science Committee (2013). Population Estimates Database, version 2013.

568

Available at http://rmbo.org/pifpopestimates. Accessed on April 26, 2015.

569

Pekarsky, S., A. Angert, B. Haese, M. Werner, K. A. Hobson, and R. Nathan (2015). Enriching

570

the isotopic toolbox for migratory connectivity analysis: a new approach for migratory

571

species breeding in remote or unexplored areas. Diversity and Distributions 21:416-427.

572
573

Post, D. M. (2002). Using stable isotopes to estimate trophic position: models, methods, and
assumptions. Ecology 83:703–718.

574

Poupin, N., C. Bos, F. Mariotti, J.-F. Huneau, D. Tome, and H. Fouillet (2011). The nature of the

575

dietary protein impacts the tissue-to-diet 15N discrimination factors in laboratory rats.

576

PlOS One 6:e28046.

577
578

Pyle, P. 1997. Identification guide to North American Birds, Part I. Slate Creek Press, Bolinas,
CA.

579

Quinlan, S. P., and D. J. Green (2011). Variation in deuterium (δD) signatures of yellow warbler

580

Dendroica petechial feathers grown on breeding and wintering grounds. Journal of

581

Ornithology 152:93-101.

582
583

R Core Team (2013). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL http://www.R-project.org/

584

Ratikainen, I. I., J. A. Gill, T. G. Gunnarsson, W. J. Sutherland, and H. Kokko (2008). When

585

density dependence is not instantaneous: theoretical developments and management

586

implications. Ecology Letters 11:184-198.

27

587

Reudink, M. W., P. P. Marra, T. K. Kyser, P. T. Boag, K. M. Langin, and L. M. Ratcliffe

588

(2009a). Non-breeding season events influence sexual selection in a long-distance

589

migratory bird. Proceedings of the Royal Society B. 276:1619-1626.

590

Reudink, M. W., C. E. Studds, T. K. Kyser, P. P. Marra, and L. M. Ratcliffe (2009). Plumage

591

brightness predicts non-breeding season territory quality in a long-distance migratory

592

songbird, the American redstart Setophaga ruticilla. Journal of Avian Biology 40:34-41.

593

Rockwell, S. M., C. I. Bocetti, and P. P. Marra (2012). Carry-Over Effects of Winter Climate on

594

Spring Arrival Date and Reproductive Success in an Endangered Migratory Bird,

595

Kirtland’s Warbler (Setophaga kirtlandii). Auk 129:744-752.

596

Ruegg, K., E. C. Anderson, K. L. Paxton, V. Apkenas, S. Lao, R. B. Siegel, D. F. DeSante, F.

597

Moore, and T. B. Smith (2014). Mapping migration in a songbird using high-resolution

598

genetic markers. Molecular Ecology 23:5726-5736.

599

Rundel, C.W., M. B. Wunder, A. H. Alvarado, K. C. Ruegg, R. Harrigan, A. Schuh, J. F. Kelly,

600

R. B. Siegel, D. F. DeSante, T. B. Smith, and J. Novembre (2013). Novel statistical

601

methods for integrating genetic and stable isotope data to infer individual-level migratory

602

connectivity. Molecular Ecology 22:4163-4176.

603

Sealy, J. C., N. J. van der Merwe, J. A. Lee-Thorp, J. L. Lanham (1987). Nitrogen isotopic

604

ecology in southern Africa: implications for environmental and dietary tracing.

605

Geochimica et Cosmochimica Acta 51:2707–2717.

606

Seber GAF (1984). Multivariate Observations. John Wiley & Sons, New York.

607

Sheehy, J., C. M. Taylor, and D. R. Norris (2011). The importance of stopover habitat for

608

developing effective conservation strategies for migratory animals. Journal of

609

Ornithology 152: 161-168.

28

610
611

Sillett, T. S., R.T. Holmes, and T.W. Sherry (2000). Impacts of a global climate cycle on the
population dynamics of a migratory songbird. Science 288:2040-2042.

612

Small, A. (1994). California Birds: status and distribution. Ibis Publishing, Temecula, California.

613

Soto, D. X., L. I. Wassenaar, and K. A. Hobson (2013). Stable hydrogen and oxygen isotopes in

614

aquatic food webs are tracers of diet and provenance. Functional Ecology 27: 535-543.

615

Still, C. J., and R. L. Powell. (2010). Continental-scale distributions of vegetation stable carbon

616

isotope ratios. In: West, J. B., G. J. Bowen, T. E. Dawson, and K. P. Tu (eds). Isoscapes:

617

understanding movements, pattern and process on Earth through isotope

618

mapping. Springer, New York, pp.179–194.

619

Stouffer, P. C., G. M. Dwyer (2003). Sex-biased winter distribution and timing of migration of

620

Hermit Thrushes (Catharus guttatus) in Eastern North America. Auk 120:836-847.

621

Studds, C. E., T. K. Kyser, and P. P. Marra (2008). Natal dispersal driven by environmental

622

conditions interacting across the annual cycle of a migratory songbird. Proceedings of the

623

National Academy of Sciences USA 105:2929-2933.

624

Stutchbury, B. J., S. A. Tarof, T. Done, E. Gow, P. M. Kramer, J. Tautin, J. W. Fox, V.

625

Afanasyev (2009). Tracking long-distance songbird migration by using geolocators.

626

Science 323:896–896

627
628
629
630

Summers, K., and M. Gebauer (1995). Status of Vaux’s Swift in British Columbia. Wildlife
Working Report WR-67. Ministry of Environment, Victoria, BC.
Székely, G. J., M. L. Rizzo (2005). Hierarchical clustering via joint between-within distances:
extending Ward’s minimum variance method. Journal of Classification 22:151-183.

29

631

Van Wilgenburg, S. L., and K. A. Hobson (2011). Combining stable-isotope (δD) and band

632

recovery data to improve probabilistic assignment of migratory birds to origin. Ecological

633

Applications 21:1340-1351.

634
635

Varmuza, K., and P. Filzmoser (2009). Introduction to Multivariate Statistical Analysis in
Chemometrics. CRC Press, Boca Raton, FL.

636

Veen, T., M. B. Hjernquist, S. L. Van Wilgenburg, K. A. Hobson, E. Folmer, L. Font, and M.

637

Klassen (2014). Determining the winter grounds of hybrid flycatchers using a multi–

638

isotope (δ2H, δ13C, δ15N) assignment approach. PLOS 9:e98075.

639

Voigt, C. C., D. Lehmann, and S. Greif (in press). Stable isotope ratios of hydrogen separate

640

mammals of aquatic and terrestrial food webs. Methods in Ecology and Evolution. DOI:

641

10.1111/2041-210X.12414

642

Voigt, C. C., K. Schneeberger, and A. Luckner (2013). Ecological and dietary correlates of stable

643

hydrogen isotope ratios in fur and body water of syntopic tropical bats. Ecology 94:346–

644

355.

645
646

Ward, J. H. (1963). Hierarchical grouping to optimize an objective function. Journal of the
American Statistical Association 58:236-244.

647

Wassenaar, L. I., and K. A. Hobson (2000). Stable-carbon and hydrogen isotope ratios reveal

648

breeding origins of red-winged blackbirds. Ecological Applications 10:911–916.

649

Webster, M. S., P. P. Marra, S. M. Haig, S. Bensch, and R. T. Holmes (2002). Links between

650

worlds: unraveling migratory connectivity. Trends in Ecology and Evolution 17:76-83.

651
652

Webster, M. S., and P. P. Marra (2005). The importance of understanding migratory
connectivity. In: Birds of Two Worlds: The Ecology and Evolution of Temperate-

30

653

Tropical Migration Systems. R. Greenberg and P. P. Marra (Eds). Johns Hopkins

654

University. Pp 199-209.

655

Wilson, S., S. L. LaDeau, A. P. Tøttrup, and P. P. Marra (2011). Range-wide effects of breeding-

656

and nonbreeding-season climate on the abundance of a Neotropical migrant songbird.

657

Ecology 92:1789–1798.

658
659
660

Wunder, M. B. (2012). Determining geographic patterns of migration and dispersal using stable
isotopes in keratins. Journal of Mammalogy 9:360-367.
Wunderle, J. M., D. Currie Jr., E. H. Helmer, D. N. Ewert, J. D. White, T. S. Ruzycki, B.

661

Parresol, and C. Kwit (2010). Kirtland’s warblers in anthropogenically disturbed early-

662

successional habitats on Eleuthera, the Bahamas. Condor 112:123-137.

663
664
665

31

666
667
668

TABLE 1. Mean (±SD) stable isotope ratios of Vaux’s Swift claws partitioned into two or three
clusters based on k-means or hierarchical agglomerative clustering, respectively. Cluster
numbers (1-3) corresponds to those shown in Figure 1.
Cluster no.
1
2
3

Two clusters
δ13C
-36.5 (11.6) -21.8 (0.4)
-42.0 (13.8) -20.9 (0.6)
δ2H

δ15N

7.0 (0.5)
8.4 (0.9)
-

Three clusters
δ13C
-45.6 (8.9) -21.8 (0.4)
-38.7 (12.3) -20.7 (0.6)
-26.3 (5.6) -21.7 (0.4)
δ2H

δ15N

7.0 (0.6)
8.6 (0.9)
7.1 (0.5)

669

32

670

FIGURE LEGENDS

671
672

FIGURE 1. Variation in Vaux’s Swift claw stable isotopes of hydrogen (δ2H: a, b), carbon

673

(δ13C: c, d), and nitrogen (δ15N: e, f) when claws were clustered into 2 (a, c, e) or 3 (b, d, f)

674

clusters. Shown are the median, interquartile range (IQR), and outliers (> 1.5 x IQR).

675
676

FIGURE 2. Clustering of stable isotope ratios of Vaux’s Swift claws into two (left panel) or

677

three (right panel) clusters based on k-means or hierarchical agglomerative clustering,

678

respectively.

679
680

33

681

FIGURE 1

682

34

683

FIGURE 2

684
685

35