1

Population response to environmental productivity throughout the annual

2

cycle in a migratory songbird

3
4

Andrew G. Pillar • Scott Wilson • Nancy J. Flood • Matthew W. Reudink

5
6
7
8
9

A. G. Pillar • N. J. Flood • M. W. Reudink (mreudink@tru.ca)

10

Department of Biological Sciences,

11

Thompson Rivers University, Kamloops, BC, Canada

12
13

S. Wilson

14

Environment Canada, National Wildlife Research Centre,

15

Ottawa, ON, Canada

16
17
18
19
20

21

Abstract

22
23

Environmental factors affect migratory animal populations in every phase of their annual cycle

24

and have significant impacts on breeding success and survival. The Breeding Bird Survey

25

provides a long-term database for examining population trends in North American birds,

26

allowing us to examine large-scale environmental factors that influence population abundance.

27

We examined plant productivity as measured by NDVI (Normalized Difference Vegetation

28

Index) over a 24 -year period from 1983 – 2006 in bird conservation regions (BCRs) that

29

overlapped Bullock’s oriole (Icterus bullockii) breeding, moult, and wintering ranges to ask

30

whether plant productivity in one year influences population abundance in the subsequent

31

breeding season. Bullock’s orioles have a moult-migration strategy, with a stopover moult in the

32

Mexican monsoon region, which necessitates examining each stationary phase of the bird’s

33

annual cycle to understand the impacts of environmental factors on population abundance. Our

34

results show increased breeding abundance in three (Great Basin, Coastal California and

35

Shortgrass Prairies) of the six BCRs in which the species breeds following years with high NDVI

36

values. We did not detect a response of breeding abundance to high NDVI values in the previous

37

year in either the moulting region or in their primary over-wintering area in central Mexico. Our

38

results demonstrate that large-scale annual variation in primary productivity on the breeding

39

grounds can have an impact on breeding abundance in the following season, but further studies

40

on migratory connectivity and on ecological mechanisms during the non-breeding seasons are

41

needed to understand why we did not detect an influence of productivity during these periods.

42
43

Keywords

44
45

Abundance • BBS • Bullock’s oriole • Climate • Long-term • NDVI

46
47

Introduction

48
49

Throughout the annual cycle, numerous biotic and abiotic factors interact to influence variation

50

in the survival and breeding success of animals. For migratory birds, studying these interactions

51

can be challenging due to the vast distances individuals travel in a year. Most studies have

52

focused on birds on their breeding grounds, at migratory banding stations, or occasionally on

53

their wintering grounds, but relatively little is known about how these phases of the annual cycle

54

interact to influence individual and population level processes (Marra et al. 1998; Webster et al.

55

2005; Bowlin et al. 2010; Harrison et al. 2013; Senner et al. 2014). A more comprehensive,

56

annual-cycle approach to the study of migratory birds, which integrates the phases of breeding,

57

moult, migration, and over-wintering, is necessary to understand the interactions among them

58

(Marra et al. 1998; Bowlin et al. 2010; Newton 2011; Harrison et al. 2013).

59

Variation in climatic conditions and the influence of this variation on resource

60

availability can have important effects on individuals throughout the annual cycle. For example,

61

spring migration of individual red-backed shrikes (Lanius collurio) and thrush nightingales

62

(Lanius luscinia) is delayed by the need for a longer stopover during periods of drought in

63

eastern Africa (Tøttrup et al. 2012). Resource availability can be measured in a number of

64

different ways, including the use of remotely sensed measures of plant productivity such as

65

NDVI (Normalized Difference Vegetation Index). In bobolinks (Dolichonyx oryzivorus),

66

declines in primary productivity, as measured by NDVI, appear to lead to early departure during

67

fall migration as well as movement from one stopover site to another (Renfrew et al. 2013).

68

Similarly, greater plant cover is positively related to diurnal body mass gain during migration

69

stopover for forest dwelling willow warblers (Phylloscopus trochilus) and Eurasian redstarts

70

(Phoenicurus phoenicurus) (Ktitorov et al. 2008). Thus, plant productivity appears to be critical

71

for migratory birds and can be measured both directly (i.e., in the field), as well as indirectly,

72

through the use of metrics such as NDVI.

73

Although climatic conditions and plant productivity may influence individuals

74

immediately and directly, these effects may also carry-over to subsequent seasons. Carry-over

75

effects occur when conditions in one phase of the annual cycle have a significant effect on the

76

animal in a subsequent phase; these effects are thought to be especially strong among migratory

77

species, and may have a significant impact on their life history and evolution (Inger et al. 2010).

78

For example, environmental conditions on over-wintering territories can carry over to impact

79

arrival dates (Marra et al. 1998; Reudink et al. 2009) and reproductive success in the following

80

season (Saino et al. 2004; Reudink et al. 2009). Annual variability in amount of rainfall in the

81

wintering range of American redstarts (Setophaga ruticilla) is associated with changes in both

82

the physical condition of individuals and the timing of their departure during spring migration

83

(Studds and Marra 2005). In the case of birds that utilize stopover sites to moult during migration

84

(i.e., moult-migrants), it is possible that these types of carry-over effects are likely to be

85

contingent on both the over-wintering environment and the condition of the stopover moult

86

location.

87

In addition to the effects of climate and environmental conditions on individuals, large-

88

scale climate patterns can have major impacts on population-level dynamics. For example, the El

89

Nino Southern Oscillation (ENSO) influences rainfall patterns and temperature and thus food

90

availability for black-throated blue warblers (Dendroica caerulescens) wintering in Jamaica,

91

leading to lower survival rates in drier, El Nino years (Sillett et al. 2000). Climate indices for

92

ENSO and other large-scale cyclical climatic events have also been shown to account for the

93

majority of variability in reproductive success of 10 other migratory bird species (Nott et al.

94

2002). Temperature trends have affected yellow-billed cuckoos (Coccyzus americanus), as

95

populations have increased in size following cooler years (Anders and Post 2006). In addition,

96

Saino et al. (2004) and Wilson et al. (2011) demonstrated that plant productivity, as measured by

97

NDVI, is a significant driver of population dynamics in both barn swallows (Hirundo rustica)

98

and American redstarts (Setophaga ruticilla). In both species, years with low NDVI on the

99

wintering grounds resulted in a significant decrease in abundance during the subsequent breeding

100

season (Saino et al. 2004; Wilson et al. 2011).

101

Although a vast majority of migratory songbirds have two stationary phases during the

102

annual cycle, some western North American songbirds, including Bullock’s orioles (Icterus

103

bullockii), have an interesting migratory strategy that includes a prolonged stationary moult

104

period in the Mexican monsoon region (Rohwer et al. 2005; Pyle et al. 2009). Rather than

105

undergoing a partial or complete post-breeding moult on the breeding grounds, birds that employ

106

this strategy make use of a stopover location to moult while en route to their wintering grounds

107

(Rising and Williams 1999; Newton 2011). These moult-migrants generally stop in the

108

Southwestern United States and Northwestern Mexico during the late summer through early fall

109

period, apparently timing moult to coincide with an increase in food abundance resulting from a

110

dramatic increase in plant productivity during and following late summer monsoon rains in these

111

areas (Rohwer and Manning 1990; Rohwer et al. 2005; Pyle et al. 2009). Because moult and

112

migration are energetically costly, these stopover sites may be crucial, allowing birds to build up

113

important fat stores, and grow high quality plumage, which is necessary for both flight and

114

communication (Hutto 1998; Leu and Thompson 2002). To our knowledge, no research to date

115

has investigated environmental conditions during each stationary phase of the annual cycle to

116

explore population-level effects of environmental variability on moult-migrants.

117

Using publically available Breeding Bird Survey (BBS) and environmental data (i.e.,

118

NDVI), we ask whether environmental variation during the three distinct stationary phases of the

119

annual cycle influences Bullock’s oriole breeding abundance in subsequent years. Bullock’s

120

orioles breed in riparian areas of Western North America from Northern Mexico to the Southern

121

Interior of British Columbia. Bullock’s orioles undergo an entire pre-basic moult while en route

122

to their over-wintering grounds in the fall (Rohwer and Manning 1990; Pyle 1997; Rising and

123

Williams 1999); After Hatch Year (AHY) birds appear to complete an entire pre-basic moult,

124

while Hatch Year (HY) birds moult contour feathers and a variable number of flight feathers

125

(Pyle 1997; A.G. Pillar et al., unpublished data). With the exception of some individuals that are

126

resident in southern California, orioles winter from Western Mexico to as far south as Costa

127

Rica, with the core of the wintering range being in Central Mexico (Rising and Williams 1999),

128

although details of migratory connectivity are currently unknown. As we expected that increased

129

plant productivity would enhance resource abundance for orioles, we predicted that increased

130

productivity in the monsoon region would be positively related to breeding bird abundance in the

131

subsequent year. Similarly, we predicted positive relationships between the NDVI in oriole

132

breeding and over-wintering regions and abundance in the next breeding season.

133

134

Material and methods

135
136

Survey data

137
138

We obtained range-wide BBS data for Bullock’s orioles from 1983 to 2006 and classified data

139

by Bird Conservation Regions (BCRs). BCRs capture a combination of geophysical and

140

environmental conditions within a specified area (Commission for Environmental Cooperation

141

1997), and are a useful way to delineate regions within which individuals would be exposed to

142

broadly similar habitat and climatic conditions (Sauer et al. 2003; Wilson et al. 2011). Bullock’s

143

orioles are present on BBS routes in 17 BCRs across western North America, but many have few

144

detections of this species, either because the BCR has few BBS routes or because it is located on

145

the periphery of the species’ range. We therefore limited our analysis to six BCRs (5, 9, 10, 16,

146

18 and 32), which represent the core of the breeding range and contained 80 % (n = 35,020) of

147

the Bullock’s orioles detected over the 24 years of BBS data used in this analysis. Within these

148

six BCRs (hereafter referred to as ‘strata’), many BBS routes only had data for a small number of

149

years or had very few Bullock’s oriole detections. To limit the influence of these routes, we

150

further restricted the analysis to routes that had been surveyed for 14 or more of the 24 years and

151

where 2 or more Bullock’s orioles were recorded on average across all years.

152
153

Normalized Difference Vegetation Index

154
155

We obtained bimonthly Normalized Difference Vegetation Index (NDVI) data from the GIMMS

156

(Global Inventory Mapping and Mapping Studies) remotely-sensed raster image dataset for a 25

157

year period from 1982 to 2006 (Carroll et al. 2004). The NDVI is calculated as (NIR-

158

Red)/(NIR+Red) with NIR (near-infrared) and Red being equal to the amount of near-infrared

159

and red light, respectively, reflected by a surface and recorded by remote sensing (Jensen 2007;

160

Pettorelli et al. 2011). The NDVI is positively correlated with several measures of plant

161

productivity including photosynthetic capacity, leaf area index, canopy extent and carbon

162

assimilation (Myneni et al. 1995; Hicke et al. 2002; Wang et al. 2005; Pettorelli et al. 2011). As a

163

measure of plant productivity, the NDVI is often used in wildlife studies as an index of resource

164

abundance, and variation in that abundance, over time or space (Pettoretti et al. 2005; Boone et

165

al. 2006; Wilson et al. 2011; Tøttrup et al. 2012). We were interested in examining how annual

166

variation in plant productivity, as measured by the NDVI, during the breeding, moult and over-

167

wintering periods influenced BBS abundance on the breeding grounds in the following season.

168

Bullock’s orioles exhibit strong breeding site fidelity among years (mean adult dispersal

169

of 111 m; Rising and Williams 1999) and therefore strata-level measures of productivity during

170

the breeding season in year t-1 were predicted to influence abundance in the following breeding

171

season through positive effects on reproductive output. The onset of breeding for Bullock’s

172

orioles is delayed with increasing latitude and decreasing longitude (i.e., more northern and

173

eastern populations within the range breed later; Rising and Williams 1999). To accommodate

174

this in our measure of productivity during the pre-breeding and breeding periods, we used NDVI

175

values averaged from April 1 – June 30 for BCRs 5, 9, 10, 16 and 18, and March 15 – June 15 in

176

BCR 32. Using Esri’s ArcGIS (Zonal Statistics) we extracted NDVI values from the raster image

177

dataset for each BCR in which these values intersected with BCRs and the known Bullock’s

178

oriole range. In each area, we excluded NDVI values for which (i) data was missing, (ii) the

179

values may have been influenced by snow reflectance, or (iii) the values were retrieved from

180

spline interpolation or average seasonal profile.

181

Because of absolute differences in the average NDVI across western North America, we

182

standardized it within each stratum. Breeding season primary productivity was very low during

183

the 1999 La Niña event, particularly in the Great Basin and Coastal California strata

184

(standardized NDVI values = -3.19 and -3.01 respectively). Large-scale climatic oscillations like

185

the ENSO influence a variety of ecosystem processes at large spatial and temporal scales

186

(Stenseth et al. 2003). Because of the potential for strong influence of this single year anomaly,

187

we ran models with and without BBS abundance data from 2000 to examine how this year

188

affected the beta coefficient and we report both in the results text for these strata. It is possible

189

that density-dependent competition and habitat quality influence inter-annual movement among

190

regions (Johnson and Grier 1988; Rohwer 2004; Studds et al. 2008), although the scale of this

191

movement is not well understood for passerines. Such an effect could mean that BBS abundance

192

is also influenced by current year productivity; however, we chose not to test this effect here

193

because fidelity to previous breeding areas appears to be relatively strong among adult Bullock’s

194

Orioles (Rising and Williams 1999) and adults likely comprise the majority of detections on BBS

195

routes. Without an ability to separate yearlings and adults, we were unable to properly test

196

whether current year conditions have an effect on settlement, especially at the scale of a BCR.

197

Movement at the scale of routes within BCRs might be more reasonable, even among adults, but

198

we did not have route level measures of NDVI to assess this.

199

Identification of the optimal temporal and spatial window for the moult and winter

200

periods is more difficult because individuals are not directly monitored and their presence in an

201

area during these periods must be inferred through other methods. The monsoon region as

202

described by Comrie and Glenn (1998) covers the Southwest United States (primarily eastern

203

Arizona and New Mexico) and western Mexico, including the states of Sonora, Chihuahua,

204

Sinaloa, Durango, Nayarit and northern Jalisco (Fig. 1). However, the moulting area for

205

Bullock’s Orioles is more concentrated in the west central portion of this region (Navarro-

206

Sigüenza et al. 2003; Rohwer et al. 2009; Atlas of Mexican Birds, unpublished data) and

207

primarily falls within two bird conservation regions: Sierra Madre Occidental (BCR 34) and

208

Planicie Costera, Lomerios y Coñones de Occidente (BCR 43). We thus defined these two BCRs

209

as the moult region and for each year, used the area weighted average NDVI values for the

210

species’ moult period (mid-July through mid-October) as described in Rising and Williams

211

(1999).

212

The Bullock’s oriole wintering range is concentrated in central Mexico, including both

213

coasts and higher elevation regions in the interior (Rising and Williams 1999). The northern and

214

southern extents of the wintering range include a narrow strip along the north Pacific coast and

215

the Chiapas region. Analyses relating NDVI to breeding abundance become statistically limiting

216

as the spatial correlation of the NDVI weakens across the area of interest. To reduce this effect

217

we focused on the core of the wintering range described by Rising and Williams (1999) with

218

further observational support from the eBird database (Sullivan et al. 2009). Although coarse, the

219

latter shows that Bullock’s orioles are most common in winter along Pacific coastal and interior

220

regions between approximately 15.7° and 20.5° North latitude and -96.6° and -105.7° West

221

longitude. This region includes BCRs 45, 46, 47, 50, 51, 53 and 54, among which the NDVI

222

shows high spatial autocorrelation: average pairwise correlation, rp = 0.62. The pairwise

223

correlations for NDVI are lower when these BCRs are compared to those in the Sinaloa region of

224

the north Pacific coast (BCRs 43, 44), the Caribbean coastal regions in Veracruz state (BCRs 49,

225

52 and 55), and those in Chiapas state at the southern extent of the species’ range (primarily

226

including BCRs 49, 52 and 58-61). Because these three regions represent the outer portions of

227

the wintering range, and because the spatial correlations of the NDVI were lower, we did not

228

include NDVI values from these regions in our analysis. To identify an annual winter NDVI

229

covariate (Dec through Mar), we calculated the average NDVI corrected for area for BCRs 45,

230

46, 47, 50, 51, 53 and 54 and used this as our estimate of winter productivity. The NDVI values

231

for the moulting and over-wintering regions were standardized prior to analysis.

232
233

Statistical analysis

234
235

Population change for Bullock’s Orioles was modeled with a hierarchical, over-dispersed

236

Poisson regression and analyzed under a Bayesian framework with Markov Chain Monte Carlo

237

(MCMC) methods in WinBUGS 1.4.3 (Lunn et al. 2000). Each count Ci,j,t (for stratum i,

238

observer-route combination j and year t) was modeled as a Poisson random variable with mean

239

λi,j,t defined as a log-linear relation to the predictor variables, which include process and

240

sampling components:

241
242

[1]

log (λi,j,t) = αi + β1,i × year + β2,i × breed NDVIt-1 + β3,i × moult NDVIt-1 + β4,i × winter
NDVIt + β5 × start-upj,t + observer-routej,t + εi,j,t

243

The model includes strata-specific estimates for average BBS abundance (αi), temporal trend

244

(β1), and the NDVI-based estimates of productivity for the breeding period in year t - 1 (β2),

245

moulting period in year t - 1 (β3) and the wintering region in year t (β4). Betas 1-4 were assigned

246

flat, normal distributions with mean 0 and variance 103. Parameter estimates at the stratum-level

247

need to incorporate sources of variation in the data associated with routes and observers (Link

248

and Sauer 2002, 2007). Observer-route combinations (ω) were treated as normal random

249

variables. Observers also differ in their experience at identifying species on BBS routes, with

250

variation being particularly high between their first year and all subsequent years (Link and

251

Sauer 2007). We incorporated these start-up effects by assigning a 1 if that count (Ci,j,t) was the

252

first year of service for that individual and a 0 otherwise. Beta 5 (β5) represents this first year

253

influence in the model and was assigned a flat, normal distribution with mean 0 and variance 103.

254

The component εi,j,t helps accommodate over-dispersion, which is common in count data such as

255

that provided by the BBS. Because of our interest in the effects of an annually-varying

256

covariate, we did not include the variance of year effects as is typically done for the analysis of

257

temporal trends using BBS data (Sauer and Link 2011). Variance associated with observer-route

258

combinations (σ2ω) and over-dispersion (σ2ε) were assumed to be constant across strata and were

259

assigned vague inverse-gamma prior distributions with shape and scale parameters = 0.001.

260

We chose not to include density dependence in our models and instead used an approach

261

similar to other analyses relating environmental covariates to BBS abundance of a species

262

(LaDeau et al. 2007; Link and Sauer 2007; Wilson et al. 2011). While abundance on BBS routes

263

in year t may be related to abundance in year t + 1, estimation of temporal autocorrelation is

264

difficult in these models. Because of missing data, this approach requires deriving an estimate for

265

abundance on a route when it was not surveyed in year t and then using that value as a predictor

266

in year t + 1. A specific estimate of the density dependent coefficient can also be biased because

267

the large sampling variation present in citizen science data makes it difficult to estimate the

268

return to an equilibrium population size (Lillegård et al. 2008). We acknowledge that temporal

269

autocorrelation may occur in our data, but there was no evidence of temporal autocorrelation in

270

the NDVI for any period or region and therefore, autocorrelation in the count data should at most

271

only influence our ability to detect a relationship between the NDVI and subsequent abundance.

272

We ran three Markov chains for the model, each for 50,000 iterations, and examined

273

convergence through individual parameter history plots, Gelman-Rubin diagnostics, and the

274

estimation of MC error/SD for each parameter (Gelman et al. 2004; Link and Barker 2010). The

275

chains typically converged within 1,000 iterations, but we discarded the first 10,000 iterations as

276

a burn-in before drawing samples from the posterior distribution. Graphical posterior predictive

277

checks were used to examine the fit of the model (Gelman et al. 2004). This measure involves

278

drawing replicate samples of data from the posterior distribution and comparing the fit of the

279

observed data to the replicate data that conform to the assumptions of the model. We examined

280

the temporal trend and NDVI influence in each period by examining the median estimate along

281

with 95 % credible intervals (CI) for β1 and β2-4 respectively. Parameters were interpreted as

282

significant when 95 % CIs excluded zero.

283

In our examination of the relationship between NDVI and BBS abundance, we performed

284

an initial run of the model to first identify significant relationships between the NDVI in a period

285

of the annual cycle and BBS abundance. We then ran the model again and in this second run,

286

included a derived estimate of the expected change in BBS abundance in relation to a 1-unit

287

change in the standardized NDVI for those strata in which the effect was significant for a period

288

of the annual cycle. The use of a standardized NDVI value in this case was because of variation

289

in the absolute NDVI among strata. The predicted stratum-specific abundance in relation to

290

NDVI was specified as a derived parameter in the model following Sauer and Link (2011) as:

291

[2]

Ni,t = exp(αi + β1,i × yeart-mid + β2,i × breed NDVIt-1 + β3,i × moult NDVIt-1 + β4,i × winter

292

NDVIt + 0.5 × σ2ω + 0.5 × σ2ε)

293

We estimated how BBS abundance varied in relation to a change in the NDVI by

294

comparing the predictions from two models that differed in one unit of the standardized NDVI

295

during that period of the annual cycle while holding other variables constant. We set the NDVI

296

beta parameters for the other periods of the annual cycle equal to 0, which is equivalent to

297

average productivity in those regions. We used the mid-point (t-mid = year 12) of the time series

298

for each stratum to incorporate temporal trends. Variance components for observer-route effects

299

and over-dispersion were included to improve estimates of mean abundance from the log-normal

300

distribution (Sauer and Link 2011).

301

We also considered the power to detect a significant relationship at 95 % certainty

302

between the NDVI and subsequent BBS abundance. This analysis was performed for each

303

stratum in relation to NDVI in the breeding, moulting and wintering regions. Power was

304

measured by taking the lower credible interval half-widths (2.5 %) of the posterior distribution

305

for β2 through β4 to predict the change in abundance for a 1 unit change in the NDVI. These half-

306

widths are the values that would be required for the credible intervals to exclude zero, assuming

307

a positive effect of the NDVI in each period of the annual cycle; they thus provide a measure of

308

the extent to which the NDVI would need to influence BBS abundance in order for us to declare

309

the existence of a significant effect in the face of other sampling and process variance.

310
311

Results

312
313

Survey data

314
315

We included BBS data from 161 routes across six strata on which 24,689 individual orioles were

316

recorded between 1983 and 2006. Bullock’s oriole abundance was highest in Coastal California

317

(BCR32), Shortgrass Prairie (BCR 18), and Great Basin (BCR9), and lower in the Northern

318

Pacific Rainforest (BCR5), Northern Rockies (BCR10 and Southern Rockies-Colorado Plateau

319

(BCR16, Table 1). Because the species was less abundant in the latter three strata, the proportion

320

of BBS routes that met our minimum criteria was lower, particularly for the Northern Rockies

321

(8.4 % of routes) and Southern Rockies-Colorado Plateau (11.5 %).

322

Population model results

323

Our model included three NDVI covariates for the breeding season, moulting region and

324

over-wintering region (Fig. 1). There was no evidence that productivity in the moulting region or

325

the over-wintering region influenced subsequent abundance for any of the six breeding strata (95

326

% credible intervals for all strata overlapped zero). In contrast, breeding season productivity

327

resulted in significantly greater abundance in the following year with 95 % certainty for the

328

Great Basin, Shortgrass Prairie and Coastal California (Table 2, Fig. 1). For the Great Basin, this

329

interpretation was based on the analysis without data from 2000, which followed the 1999 La

330

Niña event, when breeding season plant productivity was very low. The estimated coefficient for

331

the Great Basin was 0.065 [95 % CI (0.001, 0.129)] without BBS data for the year 2000 and

332

0.014 with the year 2000 included [95 % CI (-0.029, 0.057)], indicating a strong single year

333

influence on the slope of the relationship between breeding season NDVI and subsequent BBS

334

abundance. This result indicates that Bullock’s oriole abundance was higher than expected in

335

2000 given the low primary productivity associated with the La Niña event in 1999. There was

336

less influence of the La Niña event on the slope for Coastal California (β2 = 0.051 with 2000 vs

337

0.039 without 2000) or Shortgrass Prairie (β2 = 0.052 with 2000 vs 0.053 without 2000) and in

338

both cases the response from 1999 to 2000 was more similar to the pattern across all years. The

339

predicted percentage change in BBS abundance in response to a 1 unit change in the

340

standardized breeding season NDVI was 6.70 [95 % CI (0.00, 13.70)] in the Great Basin, 5.40

341

[95 % CI (0.10, 10.90)] in Shortgrass Prairie and 5.20 [95 % CI (0.40, 10.20)] in Coastal

342

California. There were no significant relationships between breeding season NDVI and

343

subsequent BBS abundance in the other three strata (Table 2). All three strata actually had

344

negative coefficients, but the credible intervals were wide due to the low precision (Table 3).

345

Our examination of precision considered what percent change in abundance would be

346

needed in response to a 1 unit change in the NDVI for the 95 % intervals of the coefficients to

347

exclude zero at the lower end (i.e., a positive response to NDVI). For the effect of prior breeding

348

season productivity, the percent change in abundance required for the Great Basin, Coastal

349

California and Shortgrass Prairies ranged from 4.19 to 4.97 %, whereas it was higher for the

350

other three strata, in which abundance was lower (ranging from 7.58 to 8.59, Table 3). The

351

pattern among strata was similar for the non-breeding periods, but precision was slightly lower

352

compared to the prior breeding season effect (Table 3).

353

Variance associated with the observer-route effect was 0.900 (posterior SD = 0.038) and

354

with the over-dispersion parameter was 0.331 (posterior SD = 0.012). The first-year effect was

355

negative [β6 = -0.140, 95 % CI -0.241, -0.042)], indicating that BBS observers tend to

356

underestimate the abundance of Bullock’s orioles in their first year of service. Trends were

357

significantly negative with 95 % certainty in the two coastal strata (Coastal California, Northern

358

Pacific Rainforest) and were not significant in the four interior strata (Table 1). The trends

359

reported here are based only on those BBS routes used in this analysis and may differ from those

360

in each stratum as a whole, although most are generally consistent with the broader Breeding

361

Bird Survey analysis (Sauer and Link 2011).

362
363

Discussion

364
365

Understanding how animal populations respond to large scale changes in climate has proven

366

difficult, especially for migratory species that inhabit areas separated by hundreds, if not

367

thousands, of kilometres during different times of the year and that are thus subject to different

368

climatic events. The North American BBS is an important long-term dataset that allows us to

369

investigate trends in breeding bird populations. By linking BBS information with data from

370

remotely-sensed estimates of primary productivity, we can begin to understand how the effects

371

of climate and subsequent green up, which occur at different locations during temporally distinct

372

periods, can influence breeding bird population trends. Our analysis suggests that over the past

373

25 years, population fluctuations in several regions across the Bullock’s oriole breeding range

374

are closely linked to changes in plant productivity during the preceding breeding season and to

375

our knowledge this is the first example showing this large scale effect of productivity on an avian

376

species in western North America. Contrary to our predictions, we did not observe any effects of

377

variation in productivity in the moulting area (a portion of the monsoon region) or on the

378

wintering grounds on subsequent breeding abundance

379

Our results indicate that breeding abundance of Bullock’s orioles in some regions is

380

positively influenced by an increase in plant productivity in the previous breeding season.

381

Specifically, we detected a significant increase in abundance in response to an increase in

382

primary productivity in the previous year for Coastal California, Shortgrass Prairie, and Great

383

Basin BCRs. Increased productivity may help with the energy demands associated with breeding,

384

feeding nestlings, and subsequent migration. There are two non-exclusive hypotheses that may

385

explain the pattern we observed. First, nestling productivity may be low during years of low

386

productivity, due to reduced food availability during breeding, resulting in lower recruitment in

387

the following season (Holmes et al. 1992). Species richness and abundance of insects (Haddad et

388

al. 2001), grassland arthropod abundance (Siemann 1998), and beetle abundance (Lassau and

389

Hochuli 2008) are all positively linked to plant productivity. Increased arthropod abundance

390

(Lepidoptera larvae) during the breeding season has been demonstrated to lead to increased

391

reproductive success in the same season for other avian species (Holmes et al. 1986). In a study

392

of stonechats (Saxicola dacotiae) endemic to a semi-arid island in the Canary Islands,

393

reproductive investment and clutch size were linked to increased arthropod abundance, which

394

was in turn related to earlier rainfall (Illera and Díaz 2008). Similarly, forest productivity was

395

positively related to insect abundance, which in turn was strongly and positively related to

396

fledgling success in ovenbirds (Seiurus aurocapilla) (Seagle and Sturtevant 2005). A second

397

hypothesis is that reduced plant productivity may lead to lowered physical condition of

398

hatchlings and/or adults, reducing the likelihood of surviving fall migration. For example,

399

declining productivity during the breeding season appears to have a negative influence on fitness

400

and reproductive success in European pied flycatcher (Ficedula hypoleuca) populations around

401

the Mediterranean (Sanz et al. 2003).

402

Our study found significant effects of breeding season productivity on subsequent

403

abundance of Bullock’s orioles in only three of the six strata. The three strata where no

404

significant effects were observed all have lower breeding abundance in general, and one

405

possibility is that we had limited potential to detect an effect. While our analysis of precision

406

showed that this was true to some extent, the coefficients for these three strata were all negative,

407

although only strongly so for the Northern Pacific Rainforest. It is possible that other factors,

408

such as predation, habitat availability or the strength of density-dependence, have a greater

409

influence on reproductive output and we were thus unable to detect a signal from environmental

410

productivity in the previous year. Reproductive output may be further affected by baseline

411

differences in productivity among ecosystems. For example, annual variation in productivity

412

may have a proportionately greater influence on avian demography in a drier ecosystem such as

413

the Great Basin compared to a wetter ecosystem like the Northern Pacific Rainforest. It is also

414

possible that because the Northern Pacific Rainforest is a very wet ecosystem at the periphery of

415

the Bullock’s oriole range, years with higher NDVI values might represent less favorable

416

conditions. Further study of the links among annual climate, plant productivity and avian

417

reproductive output in ecosystems of western North America are needed to better understand

418

why only some regions of the breeding range show this link between BBS abundance and the

419

NDVI in the previous breeding season.

420

Our finding that habitat quality in the previous breeding season influences subsequent

421

abundance raises the possibility of density-dependent compensation. Conditions experienced

422

during one stage of the annual cycle may carry-over to affect fitness and abundance during

423

subsequent stages (Runge and Marra 2005; Harrison et al. 2011) and this may in turn lead to a

424

sequential density-dependent response at the population level (Ratikainen et al. 2008). Bullock’s

425

Orioles, for instance, may have lower non-breeding season survival in years following higher

426

habitat quality on the breeding grounds when there is greater production of young. As such, we

427

may underestimate the effect of primary productivity on reproductive output, although

428

subsequent abundance would still reflect the combination of higher reproductive output followed

429

by density-dependent compensation. Higher breeding abundance in one year may also influence

430

abundance the following year, perhaps through settlement decisions (Ratikainen et al. 2008),

431

although we suspect that this is more likely to be observed at finer scales, such as among BBS

432

routes rather than among BCRs. With missing counts in the BBS dataset it was not possible to

433

use prior route-level abundance as a predictor for current abundance on routes. Moreover, others

434

have expressed concern about estimating density dependent responses in large scale datasets

435

where survey error affects our ability to properly measure the return to an equilibrium abundance

436

(Lillegård et al. 2008). Nevertheless, we acknowledge the likely importance of density-

437

dependent compensation when considering how environmental quality at different stages of the

438

annual cycle affects populations.

439

We found no evidence that productivity in the over-wintering area in central Mexico

440

affected subsequent breeding abundance of Bullock’s orioles. This result is similar to Wilson et

441

al. (2011) who found that BBS abundance of American redstarts in western North America was

442

not influenced by the NDVI in Mexico, where they are known to over-winter. This was in

443

contrast to the strong, positive relationship between American redstart abundance in eastern

444

North America and the annual NDVI in the Greater Antilles, where eastern birds primarily over-

445

winter. In this case, long-term studies of American redstarts in Jamaica (Marra et al. 1998; Marra

446

and Holmes 2001; Studds and Marra 2005, 2007) combined with knowledge of migratory

447

connectivity (Norris et al. 2006) helped identify the mechanism by which productivity affects

448

individuals and could translate into population level responses on the breeding grounds (Wilson

449

et al. 2011). High NDVI on the African over-wintering grounds also had a positive influence on

450

the reproductive success of European barn swallows (Hirundo rustica) showing how winter

451

productivity can also have carry over effects on the subsequent breeding season (Saino et al.

452

2004). In the case of Bullock’s orioles, we lack a clear understanding of migratory connectivity

453

between breeding and wintering populations and have only limited knowledge of winter ecology,

454

making it difficult to draw conclusions from these findings. Further study of migratory

455

connectivity, winter habitat use and ecosystem response to rainfall would aid our knowledge of

456

how Neotropical migrants of western North America are influenced by annual variation in

457

productivity on the wintering grounds.

458

Contrary to our predictions, we did not detect an influence of monsoon region plant

459

productivity on breeding abundance in the following year. As with the wintering grounds, we do

460

not have a full picture of where the birds from a specific breeding population moult, which

461

hinders our ability to detect the impact of conditions in the moult region on breeding populations.

462

Data from recently recovered light-level geolocators (A. G. Pillar et al., unpublished data)

463

indicate that birds from the same breeding population may moult in different locations within the

464

Mexican monsoon region, which makes it difficult to link breeding birds to their moult locations.

465

However, Comrie and Glenn (1998) demonstrated strong autocorrelation in temperature and

466

rainfall across the entire Mexican monsoon region, which may suggest that birds moulting

467

throughout the area experience similar environmental conditions. Moult-migrants using the

468

Mexican monsoon region are thought to take advantage of resources that are not available on the

469

breeding grounds, which experience drying out over the course of the summer (Pyle et al. 2009).

470

In contrast, the consistent productivity in the monsoon region makes it more likely that the

471

region provides sufficient resources to support a successful moult before the relatively short

472

migration from the moulting region to the over-wintering region. Thus, we may not have

473

detected any influence of conditions in the moulting grounds on breeding abundance simply

474

because environmental conditions consistently provide abundant, high quality resources, which

475

is likely a necessary condition for the evolution of this migratory strategy (Rohwer and Manning

476

1990; Rohwer et al. 2005; Pyle et al. 2009).

477

Because the environment in which moult takes place is important for the acquisition of

478

high quality plumage, which is necessary for both flight and mate attraction, rainfall and

479

productivity during moult may still have important consequences for the reproductive success of

480

individuals. Bullock’s orioles appear to moult over a fairly broad area; they have been observed

481

moulting in Arizona, Sonora, and Sinaloa (Pyle et al. 2009), suggesting that microsite variation

482

may be important. The quality of carotenoid-based ornamental plumage in Bullock’s orioles is

483

dependent on pigments obtained through the diet and therefore is likely linked to the quality of

484

habitat and the availability of food in the area in which moult takes place. Future work

485

examining whether the productivity of the moulting grounds or the timing of arrival in the area,

486

influences the feather colour of individual orioles would provide further insight into how these

487

environmental factors influence reproductive success at both an individual and population level.

488

A logical future step for this study would be to run a similar analysis on other moult-migrant

489

songbirds from Western North America, examining differences and similarities among western

490

moult-migrants. By using an inclusive approach, which takes into account breeding, stopover,

491

and over-wintering regions, we can gain a more complete understanding of how climatic factors

492

influence migratory bird populations, and thus be better equipped to deal with the potential

493

conservation challenges associated with climate change.

494
495

Acknowledgements

496
497

We would like to thank P. Marra, D. Carlyle-Moses, and D. Green for insightful comments and

498

suggestions on this manuscript. We would also like to thank R. Jovani for editing this manuscript

499

and S. Rohwer and X. Harrison for excellent and helpful reviews of an earlier version of the

500

paper. S. Rohwer and A. Navarro- Sigüenza provided data and helpful discussion critical for

501

delineating the moult region. Funding for this project was provided by a Natural Sciences and

502

Engineering Research Council Discovery Grant to M.W.R and a Natural Sciences and

503

Engineering Research Council Canada Graduate Scholarship to A.G.P.

504

References

505

Anders AD, Post E (2006) Distribution‐wide effects of climate on population densities of a

506
507
508

declining migratory landbird. J Anim Ecol 75:221-227
Boone RB, Thirgood SJ, Hopcraft JGC (2006) Serengeti wildebeest migratory patterns modeled
from rainfall and new vegetation growth. Ecology 87:1987–1994

509

Bowlin MS, Bisson IA, Shamoun-Baranes J, Reichard JD, Sapir N, Marra PP, Kunz TH,

510

Wilcove DS, Hedenström A, Guglielmo CG (2010) Grand challenges in migration

511

biology. Integr Comp Biol 50:261–279

512

Carroll ML, DiMiceli CM, Sohlberg RA, Townshend JRG (2004) 250m MODIS Normalized

513

Difference Vegetation Index, 250ndvi28920033435, Collection 4, University of

514

Maryland, College Park, Maryland, Day 289, 2003

515

Commission for Environmental Cooperation (1997) Ecological regions of North America:

516

toward a common perspective. The Commission, Montreal, QC, Canada

517

Comrie AC, Glenn EC (1998) Principal components-based regionalization of precipitation

518

regimes across the southwest United States and northern Mexico, with an application to

519

monsoon precipitation variability. Clim Res 10:201–215

520
521
522

Gelman A, Carlin JB, Stern HS, Rubin DB (2004) Bayesian data analysis. Chapman and
Hall/CRC, Boca Raton
Haddad NM, Tilman D, Haarstad J, Ritchie M, Knops JM (2001) Contrasting effects of plant

523

richness and composition on insect communities: a field experiment. Am Nat 158:17-35

524

Harrison XA, Blount JD, Inger R, Norris DR, Bearhop S (2011) Carry-over effects as drivers of

525

fitness differences in animals. J Anim Ecol 80:4-18

526

Harrison XA, Hodgson DJ, Inger R, Colhoun K, Gudmundsson GA, McElwaine G, Tregena T,

527

Bearhop S (2013) Environmental conditions during breeding modify the strength of

528

mass-dependent carry-over effects in a migratory bird. PLoS ONE 8:e77783

529

Hicke JA, Asner GP, Randerson JT, Tucker C, Los S, Birdsey R, Jenkins JC, Field C (2002)

530

Trends in North American net primary productivity derived from satellite observations,

531

1982–1998. Global Biogeochem Cy 16:2-1–2-14

532
533
534

Holmes RT, Sherry TW, Sturges FW (1986) Bird community dynamics in a temperate deciduous
forest: long-term trends at Hubbard Brook. Ecol Mongr 56:201-220
Holmes RT, Sherry TW, Marra PP, Petit KE (1992) Multiple brooding and productivity of a

535

Neotropical migrant, the Black-throated Blue Warbler (Dendroica caerulescens), in an

536

unfragmented temperate forest. Auk 109:321-333Hutto RL (1998) On the importance of

537

stopover sites to migrating birds. Auk 115:823-825

538
539
540

Illera JC, Díaz M (2008) Reproduction in an endemic bird of a semiarid island: a food-mediated
process. J Avian Biol 37:447-456
Inger R, Harrison XA, Ruxton GD, Newton J, Colhoun K, Gudmundsson GA, McElwaine G,

541

Pickford M, Hodgson D, Bearhop S (2010) Carry-over effects reveal reproductive costs

542

in a long-distance migrant. J Anim Ecol 79:974–982

543
544
545
546

Jensen JR (2007) Remote sensing of the environment: an earth resource perspective, 2nd edn.
Prentice Hall, Upper Saddle River
Johnson, DH, Grier JW (1988) Determinants of breeding distributions of ducks. Wildlife
Monogr 100:1–37

547

Ktitorov P, Bairlein F, Dubinin M (2008) The importance of landscape context for songbirds on

548

migration: body mass gain is related to habitat cover. Landscape Ecol 23:169–179

549
550
551
552

LaDeau SL, Kilpatrick AM, Marra PP (2007) West Nile virus emergence and large-scale
declines of North American bird populations. Nature 447:710-713
Lassau SA, Hochuli DF (2008) Testing predictions of beetle community patterns derived
empirically using remote sensing. Divers Distrib 14:138-147

553

Leu M, Thompson CW (2002) The potential importance of migratory stopover sites as flight

554

feather molt staging areas: a review for neotropical migrants. Biol Conserv 106:45–56

555

Lillegård M, Engen S, Sæther BE, Grøtan V, Drever MC (2008) Estimation of population

556

parameters from aerial counts of North American Mallards: a cautionary tale. Ecological

557

Applications 18:197-207

558
559
560
561

Link WA, Barker RJ (2010) Bayesian inference with ecological applications. Academic Press,
London
Link WA, Sauer JR (2002) A hierarchical analysis of population change with application to
Cerulean Warblers. Ecology 83:2832-2840

562

Link WA, Sauer JR (2007) Seasonal components of avian population change: joint analysis of

563

two large-scale monitoring programs. Ecology 88:49-55Lunn DJ, Thomas A, Best N,

564

Spiegelhalter D (2000) WinBUGS-a Bayesian modelling framework: concepts, structure,

565

and extensibility. Stat Comput 10:325–337

566
567
568
569
570
571

Marra, PP, Holmes RT (2001) Consequences of dominance-mediated habitat segregation in a
migrant passerine bird during the non-breeding season. Auk 118:92-104
Marra PP, Hobson KA, Holmes RT (1998) Linking winter and summer events in a migratory
bird by using stable-carbon isotopes. Science 282:1884–1886
Myneni RB, Hall FG, Sellers PJ, Marshak AL (1995) The interpretation of spectral vegetation
indexes. Geosci Remote Sens, IEEE Trans 33:481–486

572
573
574
575
576

Navarro-Sigüenza AG, Townsend Peterson A, Gordillo-Martînez A (2003) Museums working
together: The atlas of the birds of Mexico. Bull Br Orn Club 123A:207-225
Newton I (2011) Migration within the annual cycle: species, sex and age differences. J Ornithol
152:169–185
Norris DR, Marra PP, Bowen GJ, Ratcliffe LM, Royle JA, Kyser TK (2006) Migratory

577

connectivity of a widely distributed songbird, the American Redstart (Setophaga ruticilla).

578

Ornithol Monogr 61:14-28

579

Nott MP, Desante DF, Siegel RB, Pyle P (2002) Influences of the El Niño/Southern Oscillation

580

and the North Atlantic Oscillation on avian productivity in forests of the Pacific

581

Northwest of North America. Global Ecol Biogeogr 11:333–342

582

Pettorelli N, Vik KP, Mysterud A, Gaillard JM, Tucker CJ, Stenseth NC (2005) Using the

583

satellite-derived NDVI to assess ecological responses to environmental change. Trends

584

Ecol Evol 20:503–510

585

Pettorelli N, Ryan S, Mueller T, Bunnefeld N, Jedrzejewska B, Lima M, Kausrud K (2011) The

586

Normalized Difference Vegetation Index (NDVI): unforeseen successes in animal

587

ecology. Climate Res 46:15–27Pyle P (1997) Identification guide to North American

588

birds: A compendium of information on identifying, ageing, and sexing waterbirds,

589

diurnal raptors, and gallinaceous birds in the hand. Slate Creek Press, Bolinas

590

Pyle P, Leitner WA, Lozano-Angulo L, Avilez-Teran F, Swanson H, Limón EG, Chambers MK

591

(2009) Temporal, spatial, and annual variation in the occurrence of molt-migrant

592

passerines in the Mexican monsoon region. Condor 111:583–590

593

Ratikainen II, Gill JA, Gunnarsson TG, Sutherland WJ, Kokko H (2008) When density

594

dependence is not instantaneous: theoretical developments and management implications.

595

Ecol Lett 11:184-198

596

Renfrew RB, Kim D, Perlut N, Smith J, Fox J, Marra PP (2013) Phenological matching across

597

hemispheres in a long-distance migratory bird. Divers Distrib 19:1008–1019

598

Reudink MW, Marra PP, Kyser TK, Boag PT, Langin KM, Ratcliffe LM (2009) Non-breeding

599

season events influence sexual selection in a long-distance migratory bird. Proc R Soc B-

600

Biol Sci 276:1619-1626

601

Rising JD, Williams PL (1999) Bullock’s Oriole (Icterus bullockii). In Poole A, Gills F (eds) The

602

birds of North America, No. 416. The Birds of North America Inc., Philadelphia

603

Rohwer S (2004) Using age ratios to infer survival and despotic breeding dispersal in hybridizing

604
605

warblers. Ecology 85:423–431
Rohwer S, Manning J (1990) Differences in timing and number of molts for Baltimore and

606

Bullock’s orioles: Implications to hybrid fitness and theories of delayed plumage

607

maturation. Condor 9:125–140

608

Rohwer S, Butler LK, Froehlich DR, Greenberg R, Marra PP (2005) Ecology and demography of

609

east–west differences in molt scheduling of Neotropical migrant passerines. In Greenberg

610

R, Marra PP (eds) Birds of two worlds: the ecology and evolution of migration. Johns

611

Hopkins University Press, Baltimore pp 87-105Rohwer VG, Rohwer S, Ortiz-Ramirez

612

MF (2009) Molt biology of resident and migrant birds of the monsoon region of west

613

Mexico. Ornitologia Neotropical 20:565–584

614

Runge MC, Marra PP (2005) Modeling seasonal interactions in the population dynamics of

615

migratory birds. In Greenberg R, Marra PP (eds) Birds of two worlds: the ecology and

616

evolution of migration. Johns Hopkins University Press, Baltimore pp 375-389

617

Saino N, Szép T, Ambrosini R, Romano M, Moller AP (2004) Ecological conditions during

618

winter affect sexual selection and breeding in a migratory bird. Proc R Soc B-Biol Sci

619

271:681–686

620

Sanz JJ, Potti J, Moreno J, Merino S, Frías O (2003) Climate change and fitness components of a

621

migratory bird breeding in the Mediterranean region. Glob Change Biol 9:461-472

622
623
624

Sauer JR, Link WA (2011) Analysis of the North American breeding bird survey using
hierarchical models. Auk 128:87–98
Sauer JR, Fallon JE, Johnson R (2003) Use of North American Breeding Bird Survey data to

625

estimate population change for bird conservation regions. J Wildl Manage 67:372-

626

389Seagle SW, Sturtevant BR (2005) Forest productivity predicts invertebrate biomass

627

and ovenbird (Seiurus aurocapillus) reproduction in Appalachian landscapes. Ecology

628

86:1531-1539

629

Senner NR, Hochachka WM, Fox JW, Afanasyev V. 2014. An exception to the rule: carry-over

630

effects do not accumulate in a long-distance migratory bird. PLoS ONE 9:e86588

631

Siemann E (1998) Experimental tests of effects of plant productivity and diversity on grassland

632
633
634
635
636

arthropod diversity. Ecology 79:2057-2070
Sillett TS, Holmes RT, Sherry TW (2000) Impacts of a global climate cycle on population
dynamics of a migratory songbird. Science 288:2040–2042
Stenseth NC, Ottersen G, Hurrell JW, Mysterud A, Lima, M, Chan K, Yoccoz NG, Ådlansvik B
(2003) Studying climate effects on ecology through use of climate indices: the North

637

Altantic Oscillation, El Niño Southern Oscillation and beyond. Proc R Soc B-Biol Sci

638

270:2087-2096

639
640
641

Studds CE, Marra PP (2005) Nonbreeding habitat occupancy and population processes: an
upgrade experiment with a migratory bird. Ecology 86:2380–2385
Studds CE, Marra PP (2007) Linking fluctuations in rainfall to nonbreeding season performance

642

in a long-distance migratory bird, Setophaga ruticilla. Clim Res 35:115–122

643

Studds CE, Kyser TK, Marra PP (2008) Natal dispersal driven by environmental conditions

644

interacting across the annual cycle of a migratory songbird. Proc Natl Acad Sci USA

645

105:2929–2933Sullivan BL, Wood CL, Iliff MJ, Bonney RE, Fink D, Kelling S. (2009)

646

eBird: a citizen-based bird observation network in the biological sciences. Biol Conserv

647

142:2282-2292

648

Tøttrup AP, Klaassen RHG, Kristensen MW, Strandberg R, Vardanis Y, Lindström Å, Rahbek

649

C, Alerstam T, Thorup K (2012) Drought in Africa caused delayed arrival of European

650

songbirds. Science 338:1307–1307

651
652
653

Wang Q, Adiku S, Tenhunen J, Granier A (2005) On the relationship of NDVI with leaf area
index in a deciduous forest site. Remote Sens Environ 94:244–255
Webster MS, Marra PP, Greenberg R, Marra PP (2005) The importance of understanding

654

migratory connectivity and seasonal interactions. In Greenberg R, Marra PP (eds) Birds

655

of two worlds: the ecology and evolution of migration. Johns Hopkins University Press,

656

Baltimore pp 199–209

657

Wilson S, Ladeau SL, Tøttrup AP, Marra PP (2011) Range-wide effects of breeding-and

658

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

659

Ecology 92:1789-1798

660

Table 1 Sample size, abundance per route and temporal trend (1983-2006) for Bullock’s orioles in the six strata used in this analysis.

661

Columns under “Stratum” are the total number of routes and the average number of Bullock’s orioles detected per route for all routes

662

in the Stratum. Columns under “Analysis” are those values for the routes that met our minimum criteria and used in this analysis

663

(surveys in >14 of 24 years and mean of > 2 Bullock’s oriole detections per year). Trends were determined from the model and are

664

based only on the routes used in this analysis.
BCR Stratum Name

Stratum
Number of

Analysis

BUOR/route

routes

665
666
667

Number of

Trend

BUOR/route

routes

5

Northern Pacific Rainforest

82

2.1

18

5.7

-1.61 (-3.16, -0.09)

9

Great Basin

204

3.4

45

6.8

- 0.11 (-0.99, 0.80)

10

Northern Rockies

143

1.6

12

7.3

0.40 (-1.42, 2.28)

16

Southern Rockies - Colorado Plateau

130

1.7

15

4.2

-0.59 (-2.22, 1.00)

18

Shortgrass Prairie

112

4.0

31

6.3

-0.08 (-1.22, 1.12)

32

Coastal California

104

9.8

40

11.5

-1.92 (-2.80, -1.02)

668

Table 2 Beta coefficients and 95 % credible intervals for the effects of NDVI in the breeding, moulting and over-wintering regions on

669

the abundance of Bullock’s orioles on Breeding Bird Survey routes in the following year. 1The “Breed NDVIt-1” coefficient for the

670

Great Basin is from a model without data from 2000, since the 1999 La Niña event resulted in an outlier NDVI value that had

671

considerable influence on the slope (see text). Bolded results refer to significant responses with 95 % certainty.
Strata

672
673
674
675
676
677

Breed NDVIt-1

Moult NDVIt-1

Winter NDVIt

Northern Pacific Rainforest

-0.054 (-0.136, 0.029)

-0.036 (-0.119, 0.047)

0.008 (-0.077, 0.092)

Great Basin1

0.065 (0.001, 0.129)

0.033 (-0.016, 0.083)

-0.001(-0.054, 0.052)

Northern Rockies

-0.024 (-0.127, 0.079)

-0.007 (-0.109, 0.096)

0.006 (-0.097, 0.107)

Southern Rockies / Colorado Plateau

-0.014 (-0.093, 0.066)

-0.068 (-0.162, 0.027)

-0.011 (-0.108. 0.084)

Shortgrass Prairie

0.052 (0.001, 0.103)

0.041 (-0.020, 0.102)

0.005 (-0.058, 0.067)

Coastal California

0.050 (0.004, 0.097)

-0.015 (-0.064, 0.033)

-0.005 (-0.060, 0.050)

678

Table 3 Percent change in BBS abundance needed to declare a significant response to a 1 unit change in the standardized NDVI for

679

each period of the annual cycle in each strata (based on the model with all years included). The percent change in the table is

680

calculated using the lower 2.5 % credible interval half-width and is thus an indication of the power to detect a positive response with

681

95 % certainty.
Strata

682
683
684
685
686

Breed NDVIt-1

Moult NDVIt-1

Winter NDVIt

Northern Pacific Rainforest

7.88

7.91

8.11

Great Basin

4.19

4.88

5.16

Northern Rockies

8.59

9.72

9.73

Southern Rockies / Colorado Plateau

7.58

8.99

9.18

Shortgrass Prairie

4.97

5.92

6.10

Coastal California

4.55

4.80

5.36

687

Figure Legends

688

Fig 1 Bullock’s oriole breeding abundance in response to NDVI in the same Bird Conservation

689

Region (BCR). Black indicates a significant response at ≥ 95 % certainty and light grey indicates

690

no significant response. The Mexican monsoon region is shown in diagonal stripes. The

691

Bullock’s oriole wintering region is outlined in black, with the core wintering area shown in light

692

grey.

693
694
695