International Journal of Biometeorology

, Volume 51, Issue 5, pp 361–373

Climate and the complexity of migratory phenology: sexes, migratory distance, and arrival distributions

Authors

    • Woods Institute for the EnvironmentStanford University
  • Terry L. Root
    • Woods Institute for the EnvironmentStanford University
Original Article

DOI: 10.1007/s00484-006-0084-1

Cite this article as:
MacMynowski, D.P. & Root, T.L. Int J Biometeorol (2007) 51: 361. doi:10.1007/s00484-006-0084-1

Abstract

The intra- and inter-season complexity of bird migration has received limited attention in climatic change research. Our phenological analysis of 22 species collected in Chicago, USA, (1979–2002) evaluates the relationship between multi-scalar climate variables and differences (1) in arrival timing between sexes, (2) in arrival distributions among species, and (3) between spring and fall migration. The early migratory period for earliest arriving species (i.e., short-distance migrants) and earliest arriving individuals of a species (i.e., males) most frequently correlate with climate variables. Compared to long-distance migrant species, four times as many short-distance migrants correlate with spring temperature, while 8 of 11 (73%) of long-distance migrant species’ arrival is correlated with the North Atlantic Oscillation (NAO). While migratory phenology has been correlated with NAO in Europe, we believe that this is the first documentation of a significant association in North America. Geographically proximate conditions apparently influence migratory timing for short-distance migrants while continental-scale climate (e.g., NAO) seemingly influences the phenology of Neotropical migrants. The preponderance of climate correlations is with the early migratory period, not the median of arrival, suggesting that early spring conditions constrain the onset or rate of migration for some species. The seasonal arrival distribution provides considerable information about migratory passage beyond what is apparent from statistical analyses of phenology. A relationship between climate and fall phenology is not detected at this location. Analysis of the within-season complexity of migration, including multiple metrics of arrival, is essential to detect species’ responses to changing climate as well as evaluate the underlying biological mechanisms.

Keywords

NAOENSOClimate changeSexual differential migrationPhenology

Introduction

The phenology of bird migration has recently received intense study as part of efforts to assess the possible ecological impacts of global climatic change (see, for review, Lehikoinen et al. 2004). Studies have considered the relationship between climate variables and phenology in three different ways: first, for detection of climatic change (Parmesan and Yohe 2003; Root et al. 2003); second, for attribution of climatic change (Root et al. 2005); and third, to predict impacts upon species or groups of species (e.g., Ahola et al. 2004; Visser et al. 2004). These types of studies have given relatively little attention to the within-season details of migratory phenology. Some recent notable exceptions include a study of the total arrival distributions of three species (Sparks et al. 2005) and a consideration of sexual selection and climatic change (Moller 2004). Overall, however, studies on migratory phenology and climatic change generally lack analyses of phenological differences between sexes and among ages, comparisons between seasons at the same location, or sufficient data on enough species to make comparisons and generalizations amongst guilds.

Yet an extensive ornithological literature on migration exists that considers topics relevant to climatic change research, such as differential migration of sexes and age classes (e.g., Stewart et al. 2002; Kissner et al. 2003), the costs and benefits of arrival timing (e.g., Kokko 1999; Smith and Moore 2003), and the differential energetics of spring and fall migration strategies (e.g., Sandberg 1996; Benson and Winker 2005). Theoretical and empirical findings from this vast literature could be used to greatly enrich climatic change impact studies. We briefly address three such research questions that are particularly relevant to climatic change issues that engage the spatial, temporal, and biological complexity of migratory phenology, but, clearly, there are many more possibilities.

First, are all individuals within a migratory species affected similarly during the migratory season? Studies that have examined trends in arrival dates of Nearctic and Neotropical birds have found phenological shifts in the early season arrival of some species of North America migrants (e.g., Bradley et al. 1999; Ledneva et al. 2004), but fewer changes have been detected in the median of migration (Marra et al. 2005; MacMynowski et al., two submitted papers). The temporal metrics used to quantify arrival in phenological analyses have received limited methodological or ecological scrutiny. To date, we could find no published North American studies that have compared different migratory phases within the same dataset.

Second, how widespread are changes in fall migratory phenology and is fall phenology as strongly associated with climatic variables as spring migration? The majority of climatic change impact studies have examined spring phenology (IPCC 2001). While spring migratory phenology clearly has implications for the onset of breeding (Both and Visser 2001; Smith and Moore 2005), fall migratory phenology can also provide valuable information on species’ breeding phenology and the dispersal of juveniles (Bojarinova et al. 2002). Delays in fall migration have been documented for some species (e.g., Jenni and Kery 2003), while research in other locations has detected earlier departures from breeding grounds (Cotton 2003; MacMynowski et al., two submitted papers). Such differences indicate geographical variation in both climatic changes and species’ responses.

Third, are short- and long-distance migrants responding differently to climatic change? Is this associated with changes in the onset of migration or the rate and progression of migration? Initial research confirms the observation that geographical differences in the effects of global warming can have serious consequences for species that rely on resources spread across a hemisphere (Cotton 2003; Visser et al. 2004). Ahola et al. (2004) concluded that warmer early season temperatures advanced migration, but breeding phenology was constrained by later season temperatures on the breeding grounds. When compared to tropical and sub-tropical locations, greater warming rates at higher latitudes (IPCC 2001) can be expected to affect the migratory schedules of short- and long-distance migrants in different ways, particularly if conditions on the wintering grounds are strongly linked to migratory phenology (e.g., Gordo et al. 2005).

In this study, we investigate these questions using a phenology dataset containing records on the arrival date of nearly 25,000 birds from 1979 to 2002. For more than two decades, staff at the Field Museum of Natural History in Chicago, USA, have collected and catalogued all birds that have died by flying into expansive glass windows of McCormick Place, a large convention facility located on a flight corridor used by spring and fall migrating birds. In contrast to earlier research that only considers first arrival date or the median arrival of a species, these data allow us to examine arrival timing throughout the season and investigate differences between 11 short- and 11 long-distance migratory species. Detailed, comparative research on short- and long-distance migrants in North America has been limited thus far because the movements of many of the earliest short-distance migrant species start (early March) before many bird observatories begin spring banding.

We analyze within-season migratory timing using three temporal metrics of migration (first arrival, onset of continuous migration, median of migration) and discuss how these metrics capture different ecological aspects, and potentially different relationships, of species to climate. To evaluate associations between multiple spatial scales of climate and migratory phenology, we correlate regional temperature anomalies and two indices of continental-scale climate oscillations with arrival timing. The North Atlantic Oscillation (NAO) is well known for its climatic and ecological influence on northern latitudes of Europe and North America (e.g., Hurrell et al. 2003; Stervander et al. 2005). Though less recognized in eastern North America, El Niño-Southern Oscillation (ENSO) has also been identified as an influence on weather conditions in the Great Lakes (Rodionov and Assel 2003).

In summary, the purposes of this study are to evaluate the relationship between multi-scalar climate variables and differences (1) in timing of arrival between sexes, (2) in the total arrival distribution of species with varied life history traits (e.g., guilds, migratory distance), and (3) between spring and fall migration at the same location. The goal of this comprehensive comparative analysis is to increase our understanding of the complexities of migratory phenology both within- and between-species. Attention to the within-season complexity of migration will enhance our ability to make biological interpretations of climatic associations with arrival timing and, thus, help anticipate the impacts that climatic change could have on migratory songbird populations.

Materials and methods

Data source

We obtained records from the Field Museum of Natural History, Chicago, USA, for all birds found dead at McCormick Place from 1979 to 2000. McCormick Place is located approximately 1 km south of the Field Museum on the shores of Lake Michigan. Over this time period, a total of 126 species were collected through daily visits during spring migration (March–June), and 87 species during fall migration (August–October), for a total of 132 unique species. The database contains the species name, subspecies, date of collection, sex, body mass, age, skull fusion and fat condition. Data on the skull description and fat deposit were sporadically recorded and not useful for analysis. From the other data we generated two databases: (1) all species collected in the spring and fall from 1979 to 2000 for which the date of arrival (death) for each individual is known, and (2) data for 22 of the most abundant species in the first database, updated to spring 2002 (Table 1).
Table 1

Species

Common name

Latin name

Group

Statusa

Dietb

American Robin

Turdus migratorius

Short

Y

S, F, I

American Tree Sparrow

Spizella arborea

Short

W

S, F, I

American Woodcock

Scolopax minor

Short

T

S, F, I

Common Grackle

Quiscalus quiscula

Short

Y

S, F, I

Dark-eyed Junco

Junco hyemalis

Short

W

S, F, I

Field Sparrow

Spizella pusilla

Short

Y

S, F, I

Fox Sparrow

Passerella iliaca

Short

W

S, F, I

Hermit Thrush

Catharus guttatus

Short

B

S, I

Song Sparrow

Melospiza melodia

Short

Y

S, F, I

Swamp Sparrow

Melospiza georgiana

Short

Y

S, F, I

White-throated Sparrow

Zonotrichia albicollis

Short

W

S, F, I

Common Yellowthroat

Geothlypis trichas

Long

B

S, F, I

Gray Catbird

Dumetella carolinensis

Long

B

I

Indigo Bunting

Passerina cyanea

Long

T

I

Lincoln’s Sparrow

Melospiza lincolnii

Long

T

I, F

Northern Waterthrush

Seirus noveboracensis

Long

B

I, F

Ovenbird

Seirus aurocapillus

Long

B

I

Rose-breasted Grosbeak

Pheucticus ludovicianus

Long

T

I, F

Swainson’s Thrush

Catharus ustulatus

Long

T

S, I

Tennessee Warbler

Vermivora peregrina

Long

T

I, F

Veery

Catharus fuscescens

Long

T

I, F

Wood Thrush

Hylocichla mustelina

Long

B

S, F, I

aResidency status in Chicago: B breeding only; W winter only; Y year-round; T transient only

bS Seeds, F fruit, I invertebrates

The species analyzed include those that are not resident in the region near Chicago (transients only) as well as winter, summer, and year-round residents. Individuals of species in the latter category (e.g., Song Sparrow, Swamp Sparrow) migrate into the Chicago area from other locations in the region. It is unlikely that residents from more local populations would regularly strike the windows of McCormick Place given the absence of suitable habitat in proximity. The records of the European Starling (Sturnus vulgaris), which is resident and non-migratory, support this assumption. The distribution lacks peaks or a pattern throughout the year. This is in stark contrast to migratory species that have arrival distributions indicating that they move through the area over discrete time periods.

While museum records have not been previously used for phenological analysis, this dataset is very similar to more traditional (e.g., banding) datasets, e.g., consistent daily effort and coverage of the entire seasonal arrival distribution. In contrast to mass mortalities at radio and television towers, which are sporadic and associated with weather conditions (Crawford 1980; Kingsley and Kershner 1986), the window strikes at McCormick Place are relatively regular and continuous throughout both spring and autumn migration periods (Figs. 1 and 2). Furthermore, there are sex differential data on species for which sexes are not easily distinguished in the field. Of course, like any field location, this dataset can be affected by regional land use changes, e.g., increased urbanization, but there were no known changes in direct proximity to McCormick Place. Furthermore, birds are effectively ‘sampled’ during nocturnal migration (i.e., by striking the building) so changes in stopover habitat during the study period are not a factor. While we could detect no irregularities in seasonal coverage (e.g., bias from extreme weather events, disproportionate representation of species or sexes), a small possibility remains that an unknown factor could influence the sampling of species’ migratory distributions.
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-006-0084-1/MediaObjects/484_2006_84_Fig1_HTML.gif
Fig. 1

Spring arrival distribution and daily total of species (1979–2000). All 126 species combined. Open circle Total number of species/day; square total number birds/day

https://static-content.springer.com/image/art%3A10.1007%2Fs00484-006-0084-1/MediaObjects/484_2006_84_Fig2_HTML.gif
Fig. 2

Fall arrival distribution and daily total of species (1979–1999). All 87 species combined. Circle Total number of species/day; square total number birds/day

Species selection and summary

For this study, the spring season is defined as Julian day 60–160 (early March–early June) and fall is defined as Julian day 240–340 (late August–November). Daily collecting of birds consistently occurred during these seasonal time periods and there is little evidence for earlier or later passage of the species examined.

The full dataset of 127 species (spring) and 87 species (fall) was evaluated to identify species that met minimum criteria for species- or sex-specific analysis. To be included for spring or fall analysis, a species must have (1) a minimum of 10 birds collected during each season, (2) no more than a 3-year gap between seasons with the minimum number of birds collected, (3) at least a 10-year span of adequate data, and (4) two-thirds of the seasons must have data. Twenty-two species in the spring met the criteria. Of those, there were adequate data for males of 11 species and females of 9 species. In the fall, there were 12 species with suitable data, and half of those species have data to differentiate males and females.

We separated the species into two groups based upon their non-breeding distributions. Short-distance migrants are defined as species with the majority of their non-breeding distribution in the USA or northern Mexico. Migrants primarily originating from the Neotropics are classified as long-distance migrants. Two species’ winter range extends from the southern and/or eastern USA into the Neotropics. Hermit Thrush has widespread wintering range in the southern and eastern USA and is a partial migrant in some areas; it is classified as a short-distance migrant. The winter range of the Lincoln’s Sparrow is restricted to the south-central US and the breeding distribution is primarily in Canada; given that it is a complete migrant and the relatively greater distance between breeding and non-breeding range in eastern North America, it is classified as a long-distance migrant for this study. In the spring, there were 11 short-distance and 11 long-distance migrant species. In the fall, there were 6 short-distance and 6 long-distance migrant species. Species name, classification, and diet (Ehrlich et al. 1988) are summarized in Table 1.

Dates of arrival

We define three temporal metrics of migration to be used in all trend analyses and climate correlations: (1) first record of the season; (2) onset of continuous migration; and (3) median of the migratory period (Table 2). Each metric has its particular benefits and weaknesses and captures a different ecological aspect of migration (see Discussion). The first record is simply the date of the very first individual of a species to arrive. We defined the onset of continuous migration as the date during a species’ arrival distribution when the daily gaps between collecting individuals of a given species were no greater than three continuous days. The median of the migratory period is the date on which 50% of the individuals of a species for that season have been collected.
Table 2

Spring and fall migration

Species

Season

n

Median first

Median continuous

Median 50%

Δ(F-M)a

Female peaksb

Male peaksb

n (years)

Migr. length

SD

Dark-eyed Junco

Spring

1,101

65

69

74

16

2

2

20

38

13

Fall

1,093

273

279

300

3

1

1

20

49

20

Fox Sparrow

Spring

496

72

77

97

15

2

3

18

26

14

Fall

427

273

279

292

2

1

1

17

41

11

Hermit Thrush

Spring

421

73

82

96

8

2

2

22

25

9

Fall

671

270

276

285

3

1

1

21

32

9

Song Sparrow

Spring

3,302

80

82

91

14

3

3

22

55

19

Fall

290

278

282

294

0

1

1

19

39

20

American Tree Sparrow

Spring

247

81

87

107

16

2

2

13

28

15

Fall

155

300

304

320

nd

nd

nd

13

28

10

Swamp Sparrow

Spring

1,241

84

89

102

10

2

3

23

48

16

Fall

1,051

266

268

287

1

1

1

21

46

12

White-throated Sparrow

Spring

969

110

114

121

11

2

2

24

25

10

Fall

839

262

262

282

1

1

1

23

46

15

Lincoln Sparrow

Spring

387

121

125

130

8

1

2

21

21

9

Fall

334

257

260

274

5

2

1

21

36

10

Ovenbird

Spring

500

123

124

134

10

2

2

19

25

8

Fall

420

245

250

262

2

2

2

21

36

11

Swainson Thrush

Spring

215

124

125

133

1

1

1

16

21

9

Fall

332

245

248

257

3

1

1

20

29

9

Tennessee Warbler

Spring

162

125

131

138

nd

nd

nd

14

14

6

Fall

481

238

240

254

0

1

1

18

45

9

Common Yellowthroat

Spring

175

127

128

134

nd

nd

nd

17

24

7

Fall

195

254

261

270

2

1

1

16

34

12

Veery

Spring

336

125

130

135

8

1

2

15

19

7

Gray Catbird

Spring

150

121

125

133

1

1

2

16

17

7

Indigo Bunting

Spring

344

121

123

131

7

1

2

17

21

8

Northern Waterthrush

Spring

171

126

128

133

6

2

1

13

20

9

Wood Thrush

Spring

189

128

130

138

2

1

1

12

18

9

Field Sparrow

Spring

208

69

70

85

11

3

2

14

34

19

American Robin

Spring

149

86

99

116

nd

nd

nd

14

42

15

Common Grackle

Spring

180

97

104

110

nd

nd

nd

13

50

14

American Woodcock

Spring

165

98

99

110

nd

nd

nd

16

58

24

Rose-breasted Grosbeak

Spring

115

129

131

134

nd

nd

nd

8

14

9

aDifference (days) between seasonal peak abundance for male and female; see Methods

bNumber of within-season abundance peaks for each sex; see Methods, Fig. 3

nd Not enough data available to evaluate

Median arrival dates (Julian day) for three metrics of migration. First arrival is the date that the first individual of a species is collected. The onset of continuous migration is defined as the date during a species’ arrival distribution when the daily gaps between collecting individuals of a given species were no greater than three continuous days. The median of the migratory period is the date on which 50% of the individuals for a species for that season have been collected. Migration length is the mean length of the migratory period from the first to last record of a species.

Analyses

For each season and each arrival variable, we calculated: (1) trend of arrival over the study period (1979–2002); (2) length (days) of migration; (3) correlation with climate variables; and (4) total arrival distribution. Using the Jarque–Berra test for goodness of fit to a normal distribution (Judge et al. 1988), it can be reasonably assumed (P < 0.05) that the arrival dates are sufficiently normally distributed for all but one species (American Woodcock).

Correlation with climate variables

We calculated Pearson correlation coefficients between the three metrics of arrival (i.e., first arrival, continuous migration, and median) and two scales of climate variables for all species except the American Woodcock (see above); Spearman’s rank correlation coefficient was calculated for this species (Table 3). At the regional scale, we derived monthly regional averages of mean temperature anomalies (35–45°N, 75–100°W) from the Historical Climatology Network, i.e., temperature deviations from the 1961–1990 mean calculated on a 5° × 5° grid (Peterson and Vose 1997). We selected a regional scale temperature variable because local temperatures at McCormick Place are strongly influenced by Lake Michigan and not representative of broader spring or fall conditions in mid-latitude North America. Furthermore, use of a regional temperature eliminates any confounding effects from increasing urbanization and/or urban heat island. For the continental scale, we calculated a species’ association with the North Atlantic Oscillation (NAO) using a station-based index (http://www.cgd.ucar.edu/cas/jhurrell/indices.html) and with El Niño-Southern Oscillation (ENSO) using the Multivariate ENSO Index (MEI, http://www.cdc.noaa.gov/people/klaus.wolter/MEI/mei.html).
Table 3

Correlation between arrival and climate

 

 

First arrival

Continuous

Median

 

Sexa

NAO

Temp

NAO

Temp

NAO

Temp

Short-distance migrants

 Fox Sparrow

C

0.01

−0.63**

−0.11

−0.69**

−0.19

−0.62**

M

0.12

−0.72**

−0.39

−0.37

0.06

−0.60**

F

−0.05

−0.32

−0.18

0.11

0.07

−0.58

 Hermit Thrush

C

0.15

−0.60**

−0.26

−0.46**

0.06

−0.61**

M

−0.14

−0.42*

−0.35

−0.55**

−0.20

−0.40**

F

−0.03

−0.30

−0.13

−0.48*

−0.23

−0.48*

 Dark-eyed Junco

C

0.38

−0.69**

0.33

−0.57**

−0.04

−0.30

M

0.35

−0.56**

0.22

−0.43**

0.00

−0.55**

F

−0.29

0.16

−0.16

0.00

−0.09

−0.40*

 Song Sparrow

C

0.14

−0.62**

0.24

−0.60**

−0.25

−0.55**

M

0.21

−0.68**

0.25

−0.63**

−0.15

−0.51**

F

−0.04

−0.67**

−0.01

−0.74**

−0.10

−0.55**

 Swamp Sparrow

C

0.13

−0.56**

−0.36*

−0.32

0.20

−0.61**

M

0.32

−0.44*

−0.28

−0.59**

−0.45*

−0.52**

F

−0.29

−0.47*

−0.34

0.07

−0.28

−0.48*

 White-throated Sparrow

M

−0.23

−0.20

−0.27

−0.47*

0.04

−0.51**

F

−0.04

−0.06

0.11

0.14

−0.27

−0.26

 American Tree Sparrow

C

0.11

−0.53*

−0.10

−0.23

0.02

0.09

F

−0.12

−0.04

−0.26

−0.33

−0.15

0.04

 Field Sparrow

C

−0.07

−0.56

0.06

−0.47

0.08

−0.44

M

−0.17

−0.56**

−0.03

−0.40

−0.15

−0.31

 American Robin

C

−0.24

−0.21

0.11

−0.15

−0.03

−0.51*

 American Woodcock

C

0.18

0.06

−0.06

−0.06

0.29

0.06

 Common Grackle

C

0.38

−0.11

0.24

−0.54**

0.37

0.29

 White-throated Sparrow

C

−0.50*

0.09

−0.42*

−0.60**

−0.19

−0.48**

Long-distance migrants

 Ovenbird

C

−0.56**

0.01

−0.59**

−0.20

−0.18

−0.18

M

−0.36

0.41

−0.61**

−0.26

0.00

−0.30

F

−0.14

0.42

0.00

−0.30

−0.43*

−0.37

 Common Yellowthroat

C

−0.52*

0.34

−0.52*

0.01

−0.43

−0.48*

M

−0.53*

0.07

−0.34

−0.35

−0.54*

−0.75**

 Veery

C

−0.64**

−0.26

−0.47

−0.62**

−0.32

−0.24

M

−0.49*

−0.36

−0.57*

−0.61**

−0.56**

−0.31

 Indigo Bunting

C

−0.30

0.25

0.14

0.26

0.14

0.09

M

−0.10

0.00

0.10

0.02

0.12

−0.03

 Lincoln’s Sparrow

C

−0.40

−0.22

−0.57**

−0.24

−0.43*

−0.15

F

−0.31

−0.59**

−0.11

−0.41

−0.11

−0.42

 Gray Catbird

C

−0.49*

−0.24

0.01

0.01

−0.01

−0.10

 Northern Waterthrush

C

−0.47

−0.54*

−0.57*

−0.55*

−0.11

−0.19

 Rose-breasted Grosbeak

C

−0.45

0.15

−0.64*

0.46

−0.68**

0.20

 Swainson’s Thrush

C

−0.24

−0.25

−0.28

−0.09

0.16

0.31

 Tennessee Warbler

C

−0.03

−0.29

−0.16

−0.28

−0.40

−0.52**

 Wood Thrush

C

−0.58**

0.18

0.19

0.27

−0.63**

0.08

aC Combined sexes, M male, F female

**P < 0.05, *P < 0.10

We calculated the correlation coefficient between the arrival date and the climate variable for the month of arrival and the preceding month. Climate conditions in these months most likely have the strongest link to conditions experienced immediately before or during migration and, thus, are more informative than a single seasonal mean. If the correlation coefficients for both months are statistically significant, the value for the earliest month is retained (Table 3).

Seasonal arrival distribution

To examine the within-season timing of migrants passing through Chicago, we examine the pattern of total arrival distribution (i.e., smooth curve or multi-peaked), and evaluate differences in peak migration between sexes. Yearly species-specific data were too sparse for patterns to be identified. Patterns emerge, however, when data were aggregated across all 23 years.

For each species, the total number of individuals collected on a given Julian day for all years with data was plotted. Possible peaks in migration were identified using a “peak-picking” algorithm that searched 7-day time units for days on which the number of individuals recorded were at least 25% of the maximum daily total for the season (OriginLab Corporation 2002). Peaks resulting from a year or two of unusually high records or not part of a several-day trend were disregarded. Finally, a curve was fit to the arrival distribution using the daily totals and one to three peaks identified for each sex of each species (see, e.g., Fig. 3). Non-linear curve fitting was conducted with an unweighted Lorentz function and the Levenberg–Marquardt least-squares-fitting algorithm (OriginLab Corporation 2002). The same process was used to produce the curves representing the arrival distributions of the combined sexes for each species (Figs. 4 and 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs00484-006-0084-1/MediaObjects/484_2006_84_Fig3_HTML.gif
Fig. 3

Example of spring sexual differential migration, Swamp Sparrow. Note three within-season peaks in abundance for male and two peaks for female. Circle and solid line Male (r2 = 0.72); square and dashed line female (r2 = 0.65)

https://static-content.springer.com/image/art%3A10.1007%2Fs00484-006-0084-1/MediaObjects/484_2006_84_Fig4_HTML.gif
Fig. 4

Spring total arrival distribution. a Warblers and Indigo Bunting: solid Ovenbird (r2 = 0.86), dash Northern Waterthrush (r2 = 0.73), dot Tennessee Warbler (r2 = 0.83), dash-dot Indigo Bunting (r2 = 0.81). b Thrushes: solid Hermit Thrush (r2 = 0.84), dash Swainson’s Thrush (r2 = 0.70), dot Wood Thrush (r2 = 0.87), dash-dot Veery (r2 = 0.90). c Sparrows: solid White-throated Sparrow (r2 = 0.93), dash American Tree Sparrow (r2 = 0.78), dot Field Sparrow (r2 = 0.76), dash-dot Lincoln’s Sparrow (r2 = 0.83), short dash Song Sparrow (r2 = 0.85), short dot Swamp Sparrow (r2 = 0.92), short dash-dot Fox Sparrow (r2 = 0.87)

https://static-content.springer.com/image/art%3A10.1007%2Fs00484-006-0084-1/MediaObjects/484_2006_84_Fig5_HTML.gif
Fig. 5

Fall total arrival distribution. a Warblers and thrushes: solid Tennessee Warbler (r2 = 0.82), dash Ovenbird (r2 = 0.57), dot Common Yellowthroat (r2 = 0.70), dash-dot Swainson’s Thrush (r2 = 0.88), dash-dot-dot Hermit Thrush (r2 = 0.86). b Sparrows: solid White-throated Sparrow (r2 = 0.78), dash American Tree Sparrow (r2 = 0.77), dot Dark-eyed Junco (r2 = 0.69), dash-dot Song Sparrow (r2 = 0.59), dash-dot-dot Lincoln’s Sparrow (r2 = 0.73), short dash Swamp Sparrow (r2 = 0.77), short dot Fox Sparrow (r2 = 0.59)

In order to compare species with different relative abundances, arrival distributions were standardized by dividing the total number of birds collected on a given Julian day by the total number of birds of that species collected throughout the entire season (i.e., arrival distributions were based on percentages of total annual captures rather than absolute numbers). Individual daily data points for the two sexes of the Swamp Sparrow are plotted on Fig. 3. Including data points on the plots for each species makes comparison of the distributions among multiple species difficult. Therefore, for clarity, the subsequent Figs. (4 and 5) include only the curve fitted by the method described above.

The number of within-season relative peaks of abundance for each sex of a species is recorded in Table 2. Many studies of sexual differential migration disregard these within-season patterns and only record the difference in first arrival dates of males and females. However, since total arrival distributions are actually often multi-peaked (Table 2) and indicate peaks of varying abundance (Fig. 3), we defined the difference between male and female arrival as the difference in days between the peaks of highest abundance for each sex.

Results

Climate variables

There are no significant trends (P < 0.10) over the study period (1979–2002) for any of the spring or fall climate variables examined. NAO is uncorrelated (P > 0.10) with local or regional temperatures. As expected, local and regional temperatures are highly correlated (r = 0.55–0.81, P < 0.001).

Correlations between arrival and climate

Correlation coefficients between spring arrival dates and climate variables for each species are listed in Table 3. There were virtually no significant associations between arrival dates and climate variables in the fall. Only two species’ fall FAD (Swamp Sparrow, r = 0.61, P = 0.01; Fox Sparrow, r = 0.53, P = 0.03) were positively associated with NAO. The fall median arrival date of one species was negatively correlated with regional temperature (Song Sparrow, r = −0.46, P = 0.05).

The proportions of species, of the total in each category (males, females, combined sexes, short- and long-distance migrants) with significant correlations between spring arrival dates and climate variables are summarized in Table 4. Only two species’ first arrival dates correlate with MEI (Hermit Thrush, r = 0.57; Northern Waterthrush, r = 0.47; P < 0.05); thus, we only list NAO in Tables 3 and 4. These two correlations with MEI are likely spurious; we do not find evidence for the influence of ENSO on migratory phenology at this location.
Table 4

Proportion of species with spring climate correlations

 

All species

Long-distance migrants

Short-distance migrants

Both sexes

Male

Female

Both sexes

Male

Female

Both sexes

Male

Female

NAO

 Early phase migration

0.45

0.27

0.00

0.73

0.75

0.00

0.18

0.00

0.00

 Median arrival

0.14

0.27

0.14

0.27

0.50

0.33

0.00

0.14

0.00

 Early and median

0.14

0.18

0.00

0.18

0.50

0.00

0.00

0.00

0.00

Mean temperature

 Early phase migration

0.45

0.73

0.33

0.18

0.25

0.66

0.73

1.00

0.33

 Median arrival

0.36

0.64

0.44

0.18

0.25

0.33

0.50

0.86

0.50

 Early and median

0.23

0.55

0.33

0.00

0.00

0.00

0.45

0.86

0.33

The proportions of species with significant correlations (P < 0.10) between arrival dates and climate variables are summarized from Table 3. A summary of three temporal periods is reported: (1) early phase migration, (2) median arrival, and (3) early and median migration. The values for ‘early phase migration’ represent the total number of species with a significant correlation between a climate variable and the first arrival date and/or the date of the onset of continuous migration. The values for ‘early and median’ represent the total number of species with a significant correlation between a climate variable and both early and median arrival dates.

We identify four patterns in the correlations between species’ arrival dates and climate variables. First, as indicated in Tables 3 and 4, spring phenology is frequently correlated with climate while there are virtually no correlations between arrival dates and climate variables in the fall (see above for fall results). Second, the spring arrival of males more frequently correlates with temperature than females. Specifically, early phase migration and mean temperature are correlated significantly for males in 73% of species and for females in 33% of species (Table 4).

Third, short-distance migrant species significantly negatively correlate with temperature more frequently than long-distance migrants, but the spring arrival of long-distance migrants is negatively correlated with NAO. Seventy-three percent of short-distance migrant species (both sexes, early phase migration) correlate with temperature while this is the case for only 18% of long-distance migrants. The categories are reversed for NAO: long-distance migrant species significantly correlate with NAO for 73% of species in contrast to 18% of short-distance migrant species (Table 4). Fourth, the early phase of a species’ spring migration is correlated with climate variables, particularly NAO, much more frequently than the median of the species’ migratory passage. Forty-five percent of all species’ early migration correlates with NAO while the median of arrival correlates with NAO for only 14% of species. Temperature is more comparable, with 73% of short-distance migrant species (both sexes) correlated with the early part of the migratory period and 50% of species correlated for median arrival (Table 4).

Trends in arrival

Changes in the sampling of a species, either through changes in population size (Tryjanowski and Sparks 2001) or sampling (e.g., observation/banding effort) can affect the detection and interpretation of changes in date of arrival. If a species increases in number during the study period, earlier detection of individuals might occur and, thus, there is an apparent earlier arrival. Examination of seasonal total records, for individual species and all species combined, indicates that this is not a methodological concern for this study. The spring and fall seasonal totals of individuals of all species declined from the mid 1990s onward, with an overall (1979–2000) non-significant trend (spring, P = 0.46; fall, P = 0.35). Correspondingly, seasonal totals of individual species declined over the last third of the study period, with the Field Sparrow significantly declining over the entire spring study period (P = 0.001). These declines are consistent with research indicating overall population declines of North American songbirds (e.g., Terborgh 1989; Ballard et al. 2003).

For spring phenology, only 4 species of 22 species have a significant (P ≤ 0.05) trend in arrival for one metric of arrival (e.g., first arrival date). For fall phenology, the Dark-eyed Junco is significantly (P < 0.05) later across all three arrival metrics. There are no other significant trends in fall arrival for other species. We find little evidence for changes in spring or fall phenology at this location during the study period.

Arrival distributions

The patterns of spring and fall migration at McCormick Place are distinctly different (Figs. 1 and 2). In contrast to spring: (1) the fall migratory period is half as long as the spring migratory period; (2) half as many species occur in the fall; and (3) the total number of individuals migrating in the fall are on average only a quarter of the total recorded in the spring.

Spring and fall migratory phenology are strikingly different for individual species as well (Figs. 4 and 5). Long-distance migrants arrive later in spring and depart earlier in fall; conversely, short-distance migrants arrive earlier in spring and depart later in fall (Table 2). The spring passage of the three warbler species and of Indigo Bunting occurs during a short period in late April–mid May (Fig. 4a). Similarly, three of the four thrush species arrive in May, the exception being the short-distance migrant Hermit Thrush with a peak arrival in early April (Fig. 4b). The seven sparrow species arrive throughout the spring season, with earliest species in mid-March (American Tree Sparrow, Song Sparrow) and the latest species in May (Lincoln’s Sparrow) (Fig. 4c).

In contrast to the multi-peaked spring arrival distributions, 10 of 12 species have wide, single-peaked arrival distributions in fall. This is consistent with the lack of evidence for differential movements between the sexes of a species and more protracted fall migration periods (see below). The latest spring migrants are generally the earliest fall arrivals (Fig. 5a). Four of the five warbler and thrush species have already passed through Chicago by the time the first sparrows begin arriving in late-September (Fig. 5b). As in spring, the exceptions are the Hermit Thrush and the Lincoln’s Sparrow.

There is strong evidence of differential timing of spring migration by males and females of specific species, but little evidence of timing differences during fall migration. During spring migration, there is a difference of 5–16 days between the peak arrival of males and females for 14 of 17 (82%) species examined (Table 2). In all cases, males arrived first. The spring arrival distribution of male and female migrants, for the majority of species examined, is quite complex. For the spring migrants, females of 10 species (59%) and males of 15 species (88%) show two or more peaks in abundance in their arrival distribution (see Fig. 3: Swamp Sparrow, as an example). In contrast, we find multiple peaks for only two fall species (Lincoln’s Sparrow, Ovenbird).

Length of migration

On average, the length of spring migration is shorter than fall (spring: n = 22, mean = 29 days, SD = 13; fall: n = 12, mean = 39, SD = 7; t-test, assuming unequal variances; t = −2.61, df = 32, P < 0.01). Consequently, there is more temporal overlap amongst species in the fall compared to the spring. A possible explanation for the longer fall migration is varied times of departure from the breeding grounds (e.g., failed breeders and multiple clutches; Bojarinova et al. 2002) while there are advantages to arrival, for both males and females, as early as possible in the spring (e.g., Kokko 1999; Smith and Moore 2005).

Discussion

Our results demonstrate measurable differences in migratory phenology between seasons, sexes, and short- and long-distance migrants. These patterns correspond to species’ correlations with climate. Many of the differences are only made apparent through detailed analysis of the total arrival distribution. As we elaborate below, attention to the complexity of migratory phenology is essential for detection of species’ responses to changing climate as well as evaluating the underlying biological mechanisms.

Climate and arrival timing

Only 1 of 12 species’ fall arrival is correlated with temperature. In contrast, the spring phenology of nearly three-quarters of the 22 species examined is correlated with temperature or NAO. At McCormick Place, there is little evidence for a relationship between fall migratory phenology and the climate variables examined in this study. Thus, the remainder of our discussion focuses on spring migration.

The early migratory period (e.g., first arrival) for the earliest arriving species (i.e., short-distance migrants) and the earliest arriving individuals of a species (i.e., males) are most frequently correlated with climatic conditions. At this mid-latitude location, NAO is significantly correlated with the initial migration phase of males for nearly a third of all species examined; the arrival dates of 75% of males from long-distance migrant species correlate with NAO. The arrival dates of males for all of the short-distance migrant species are correlated with regional temperature. All in all, the arrival dates of 16 of 22 (72%) of the species in this study correlate with either NAO or regional temperature in the early phase of the species’ migratory period.

These findings strongly suggest the influence of winter/spring conditions upon migratory phenology for several North American passerine species. Research on stopover ecology indicates increasing rates of mass gain as the spring season progresses (Dunn 2000), suggesting that early season food availability may be a constraint on the timing and rate of migration for some species, or resources at wintering sites play an important role in determining migratory condition and phenology (e.g., Marra et al. 1998; Gordo et al. 2005).

While migratory phenology has been repeatedly correlated with NAO in Europe (e.g., Zalakevicius 2001; Forchhammer et al. 2002; Kanuscak et al. 2004; Stervander et al. 2005), we believe that this is the first documentation of a significant association in North America. A previous study in eastern North America examined only Neotropical migrants and only the median of arrival; no significant correlation with NAO was detected (Marra et al. 2005). In this study, NAO is negatively correlated with the early phase of migration for nearly half of the 22 species examined in this study, and particularly notable, for 8 of 11 (73%) long-distance migrant species. This latter finding suggests that large-scale climate oscillations in the western hemisphere directly or indirectly influence the onset or rate of migration for species wintering in the Neotropics, as has been documented for Palearctic–African migrants.

Proposed mechanisms of NAO influence on migratory timing include stronger winds to assist migration, improved foraging on wintering grounds, decreased migration distance by wintering farther north, and earlier development of spring (Vahatalo et al. 2004). Since local and regional temperatures at this mid-latitude site are uncorrelated with NAO, we speculate that NAO influences conditions experienced by Neotropical migrants earlier in the season, e.g., wintering grounds and southern migratory route. Supporting this explanation, in Europe, a higher NAO index has been linked to increased spring tailwinds in southern Europe and earlier migrant arrival at northern latitudes (Sinelschikova et al., submitted) as well as earlier departure from African wintering grounds due to improved foraging conditions (Forchhammer et al. 2002). Further research to identify the mechanism(s) influencing specific species or migratory routes is highly desirable, particularly in North America where the influence of NAO on migration has received less attention.

Moreover, regional temperature is correlated with the arrival of the majority of short-distance migrant species, but less than 20% of long-distant migrants at this location. Geographically proximate conditions (i.e., regional temperature as a proxy of spring phenology) are apparently more influential for the phenology of short-distance migrants while a continental-scale climate pattern (e.g., NAO) seemingly influences the migratory timing of long-distance migrant species. This is consistent with the hypothesis above that early season conditions limit the onset and/or rate of migration for some species; thus, the arrival of early (short-distance) migrant species are more frequently correlated with temperature than later season arriving (long-distance) migrants.

Quantifying migratory phenology

Our findings indicate the importance of geographic variation and small-scale seasonal changes when using phenology to detect impacts of climatic change. No species at this location demonstrated a significant trend in arrival across its arrival distribution. Two earlier studies in the region have examined spring migrant phenology: one identified spring phenological changes in many bird species (Fairfield, Wisconsin: Bradley et al. 1999) and the other did not (southern Illinois: Strode 2003). Neither study investigated the entire arrival distribution, and Strode (2003) analyzed only the phenology of warbler species. Thus, direct comparison of findings across the region is difficult.

The need for attention to the measurement of migratory phenology is evident. Depending upon how phenology is quantified, different aspects of migratory passage are captured, with the potential for different inferences on relationships between climate and species’ phenology. While NAO is significantly correlated with the initial migration phase for nearly half of the species examined in this study, the median of migration for those same species is largely uncorrelated with either ENSO or NAO. Furthermore, even though other studies report evidence of earlier migratory phenology in the Great Lakes region, it is unclear whether the timing of the entire migration period is shifting, or whether the migratory period is lengthening for some species.

The first sighting of a species (often the only data available from individual observer datasets) might be most sensitive to changes in early spring. This is supported by our finding that correlations between arrival and climate variables are much more frequent for the early phases of migration. The first arrival date is also most susceptible to outliers and, particularly for sex-differential migratory species, likely captures the arrival of males. In this sense, while sensitive to changes, first arrival indicates little about changes to the arrival distribution of a species. Furthermore, detection of trends might be disproportionately influenced by the earlier arrival of males (see Results; also Francis and Cooke 1986).

In contrast, a measure of the onset of continuous migration has the advantage of minimizing outliers and indicating the start of the migratory period for the bulk of migrants within a species. This approach has the benefit of only considering the presence–absence of a species, thereby being less susceptible to bias from days with unusually high encounters or variability in observation effort during the season. For low abundance species, however, either of these approaches may be misleading because of low frequency encounters.

The median of the migratory period captures the midpoint of a species’ passage through a location. Since it assumes consistent effort within a season, it is more susceptible to bias from inconsistent data collection effort and would not be desirable unless seasonal observation effort was certain. Furthermore, the median is less of an indicator of species’ responses to changes early in the migratory period. This might explain the reduced detection of arrival trends in North American studies using the median of migration (Marra et al. 2005; MacMynowski et al., two submitted papers) and very few significant correlations between median arrival and NAO in our study and in an earlier analysis (i.e., Marra et al. 2005).

Seasonal arrival distribution

Examination of the total arrival distribution of species further reveals the complexity of within-season migratory phenology. In the spring, multiple peaks of abundance are identified for the majority of species and males migrate earlier for 80% of species examined. As discussed above, the arrival of males more frequently correlates with climate variables than for females; there are possible implications for fitness and breeding phenology as a result (Moller 2004). The database for this analysis does not include information on age; as a result, we can only speculate whether the multiple peaks per sex are evidence of differential age-class migration. Age-class differential migration is, however, widely documented for both Nearctic–Neotropical and Paleartic–African passerines (Spina et al. 1994; Stewart et al. 2002).

Another possibility is the arrival of individuals from different parts of the wintering range. If different sexes or age classes of a species have different winter distributions, which can affect migratory phenology (Nolan and Ketterson 1990), there is the potential for these subpopulations to respond differently to geographic variations in climatic change. Thus, the within-species phenological response to climatic changes could be even more spatially and temporally complex than previously anticipated if distributional and timing differences in age and sex classes are considered.

Furthermore, different guilds tend to migrate at similar times within the season, with seed-eaters (e.g., sparrows) moving through earliest and species that are predominantly insectivores (e.g., warblers) latest. The temporal synchrony among guilds is evident when total arrival distributions are compared (Figs. 4 and 5). If there are geographic differences in the changes in spring conditions from global warming, i.e., northern latitudes warm at a greater rate compared to tropical and sub-tropical latitudes (IPCC 2001) or early season conditions change more than later season conditions (Ahola et al. 2004), species might be receiving different climatic cues across their wintering or breeding ranges. As described above, there is evidence that short- and long-distance migrants in North America respond to different climatic cues before and/or during spring migration. Glimpsing migration through a single arrival metric masks the biological richness of intra- and inter-specific phenology.

Concluding remarks

In this study, we deconstruct spring and fall phenology at a mid-latitude location in North America and consider how different aspects within and between species—sexes, migratory distance, and guilds—contribute to the complexity of migratory phenology within a season. Clearly, the geographic and biological factors that influence migratory phenology are many, e.g., winter distribution (Nolan and Ketterson 1990), availability of food (Dunn 2000), pre-departure conditions (Gordo et al. 2005), reproductive advantage (Smith and Moore 2005), and timing of molt (Schaub and Jenni 2001). Evaluating the total arrival distribution within a species (e.g., multiple-peaked sexes) and timing amongst many species (e.g., Neotropical and Neartic migrants) in relation to the direct and indirect influences of climate will help us to better understand the underlying mechanisms of migration itself, which in turn will lead us towards predicting how it might change in the future due to global warming. Given the differing reports of changes in migratory timing in the same region (this study; Bradley et al. 1999; Strode 2003), between short- and long-distance migrants at different locations (Cotton 2003; Jenni and Kery 2003), and varying changes in the arrival of different populations of the same species (Sanz 2003), detailed analyses, particularly large-scale, multi-taxa studies, are needed to elucidate the consistency, type, and magnitude of phenological changes that are occurring.

Indeed, at this point, the impact of changes in migratory phenology upon songbird populations largely remains an open question. Initial research indicates connections between migratory phenology and breeding phenology (Both and Visser 2001; Ahola et al. 2004) as well as reproductive success (Dunn 2004; Visser et al. 2004; Smith and Moore 2005). Selective pressures are certainly possible on the earliest arriving individuals (e.g., Kokko 1999), but how the bulk of the population will be affected is less obvious. If spring advances due to global warming, will all or part of the population shift their phenology? Furthermore, currently, certain species arrive concurrently (Fig. 4a,b). Given the evidence for the role of community dynamics in optimizing migration strategies (Howlett et al. 2000), what are the implications for inter-specific interactions if species’ phenology changes at different rates due to either environmental or evolutionary constraints?

Opportunities for future research—connecting the complexity of migration with the complexity of climate—are extensive and will offer both current ecological insight and a view into species’ possible responses to a warming world.

Acknowledgements

We would like to thank that numerous, unnamed individuals who collected birds at McCormick Place for 20 years so that these unfortunate avian deaths might contribute to our understanding of bird migration. Specifically at the Field Museum of Natural History, we thank Dave Willard for supplying the initial database for our investigation. Two anonymous reviewers provided constructive comments that improved this manuscript. D.P.M. was supported by a grant from the Winslow Foundation.

Copyright information

© ISB 2007