56

ARTICLE
Rapid advancement of spring migration and en route adjustment
of migration timing in response to weather during fall migration
in Vaux’s Swifts (Chaetura vauxi)

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

E.D. Prytula, A.E. McKellar, L. Schwitters, and M.W. Reudink

Abstract: Climate change has generated earlier springs, later falls, and different weather patterns. These changes may prove
challenging to migratory species if they are unable to adjust their migratory timing. We analyzed changes in migratory timing of Vaux’s Swifts (Chaetura vauxi (J.K. Townsend, 1839)) by examining first arrivals (date the first swift arrived) and peak
roost occupancy (date the maximum number of swifts were observed) at migratory roosts in both spring and fall from the
citizen science organization Vaux’s Happening. First arrivals and peak occupancy date in Vaux’s Swifts advanced over time
from 2008 to 2017, and the timing of first arrivals advanced with an increase in local wind gust speeds. In contrast, fall
migration timing did not change over time from 2008 to 2016, but higher temperatures were associated with later fall
migration (both first arrival and peak roost occupancy) and higher local wind speeds were associated with earlier fall migration (peak roost occupancy only). Like many other migratory birds, Vaux’s Swifts may be tracking earlier spring phenology,
and may also be altering their migratory timing in response to local weather conditions, especially during fall migration.
Our results indicate that swifts may be able to adjust their migration to a changing climate, at least in the short term.
Key words: Vaux’s Swift, Chaetura vauxi, citizen science, roost, migration, climate change, weather.
Résumé : Les changements climatiques ont produit des printemps plus hâtifs, des automnes plus tardifs et de nouveaux
régimes météorologiques. Ces changements pourraient poser des défis pour les espèces migratrices qui ne sont pas en
mesure d’ajuster le moment de leurs migrations. Nous avons analysé les changements du moment des migrations de martinets de Vaux (Chaetura vauxi (J.K. Townsend, 1839)) en examinant leur première arrivée (date d’arrivée du premier martinet)
et la pointe d’occupation d’aires de repos (date où le plus grand nombre de martinets est relevé) dans des aires de repos
migratoires au printemps et à l’automne, obtenues par l’organisation de science participative « Vaux’s Happening ». Les
dates de première arrivée et de pointe d’occupation des martinets de Vaux sont de plus en plus hâtives au fil du temps de
2008 à 2017, et plus les vitesses des rafales locales sont grandes, plus la première arrivée est hâtive. Le moment de la migration automnale ne change toutefois pas de 2008 à 2016, mais de plus hautes températures sont associées à des migrations
automnales plus tardives (en ce qui concerne tant la première arrivée que la pointe d’occupation) et de plus grandes vitesses
du vent sont associées à des migrations automnales plus hâtives (seulement pour les pointes d’occupation d’aires de repos).
À l’instar de nombreux autres oiseaux migrateurs, les martinets de Vaux pourraient suivre une phénologie automnale plus
hâtive et pourraient aussi modifier le moment de leurs migrations en réponse aux conditions météorologiques locales, tout
particulièrement durant la migration automnale. Nos résultats indiquent que les martinets pourraient être en mesure
d’ajuster leurs migrations aux changements climatiques, du moins à court terme. [Traduit par la Rédaction]
Mots-clés : martinet de Vaux, Chaetura vauxi, science participative, aire de repos, migration, changements climatiques, météo.

Introduction
The timing of bird migration evolved as a trade-off between
maximizing reproductive output on the breeding grounds and
minimizing mortality throughout each phase of the annual cycle
(Nathan et al. 2008; Alerstam 2011). Arrival timing on the breeding
grounds is a balance between arriving early enough to acquire a
suitable mate or territory (Cristol 1995) and avoiding harsh environmental conditions during migration and at the breeding grounds
(Brown and Brown 1998; Møller et al. 2008; Hurlbert and Liang
2012). Similarly, the timing of fall migration must balance conditions on the breeding grounds following the breeding season and

future conditions that will be experienced on the winter grounds
and en route throughout migration (Jenni and Kery 2003). With the
rapid onset of global climate change (Cook et al. 2013), a major
unknown is which species or populations will be able to migrate
earlier and time their arrival to the breeding grounds to match
the advancement of spring conditions on the breeding grounds or
adjust their migration timing in response to changes in weather
and resources experienced en route.
Climate change may affect migratory timing by altering
weather patterns, advancing spring phenology, and (or) changing
fall phenology in temperate North America (Schwartz et al. 2006).
Changes to weather patterns are predicted to alter plant and insect

Received 29 April 2021. Accepted 14 September 2021.
E.D. Prytula and M.W. Reudink. Department of Biological Sciences, Thompson Rivers University, Kamloops, BC V2C 0C8, Canada.
A.E. McKellar. Environment and Climate Change Canada, 115 Perimeter Road, Saskatoon, SK S7N 0X4, Canada.
L. Schwitters. Vaux’s Happening, Issaquah, WA 98027, USA.
Corresponding author: M.W. Reudink (email: mreudink@tru.ca).
© 2021 The Author(s). Permission for reuse (free in most cases) can be obtained from copyright.com.
Can. J. Zool. 100: 56–63 (2022) dx.doi.org/10.1139/cjz-2021-0089

Published at www.cdnsciencepub.com/cjz on 9 November 2021.

Prytula et al.

57

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

Fig. 1. Distribution of Vaux’s Swifts (Chaetura vauxi) roosts used in an analysis of spring and fall migration timing from the Citizen Science
project “Vaux’s Happening in North America”, as well as breeding, migration, year-round, and nonbreeding ranges. WGS 84 and EPSG 3857
were used in creating the map. Range data from BirdLife International and Handbook of Birds of the World (2019). Colour version online.

phenology both during migration and on the breeding grounds,
which in turn may alter the optimal timing of avian migration and
breeding (Kelly et al. 2016). Events in spring have been rapidly
advancing, with earlier plant leaf-out and flowering, and the emergence of insects occurring 2.8 days per decade earlier in the northern
hemisphere (Parmesan 2007). In response, some migratory birds
have been advancing their spring migration (Rubolini et al. 2004).
Species across a broad range of avian taxa, such as the Broad-tailed
Hummingbird (Selasphorus platycercus (Swainson, 1827)) (McKinney
et al. 2012), White Stork (Ciconia ciconia (Linnaeus, 1758)) (Gordo and
Sanz 2006), Eurasian Blackcap (Sylvia atricapilla (Linnaeus, 1758))
(Bearhop et al. 2005), and European Pied Flycatcher (Ficedula
hypoleuca (Pallas, 1764)) (Coppack and Both 2002; Both et al. 2005),
have significantly advanced their spring migration. Changes to
fall migratory timing generally appear to be weaker and more
variable across species, with short-distance migrants departing
breeding grounds later and long-distance migrants either departing
earlier or not changing departure time at all (Jenni and Kery 2003;
Brisson-Curadeau et al. 2019). In a study across Northern Europe,
80% of the variability in fall migration timing was accounted for by
weather on the breeding grounds and at stopover sites (Haest et al.
2019).
Vegetation conditions experienced en route can influence migration timing in both spring and fall (La Sorte and Graham 2021). For
example, the migratory pathways used by Painted Buntings (Passerina
ciris (Linnaeus, 1758)) are directly associated with primary productivity during fall migration (Bridge et al. 2015). Similarly, individual
Barnacle Geese (Branta leucopsis (Bechstein, 1803)) time their spring
migration to a “green wave” of primary productivity and arrive
at stopover sites when food resources peak (Si et al. 2015). Direct
effects of weather conditions experienced en route, such as temperature and wind, can also alter migratory timing (Bozó et al.
2018). Studies have documented changes in the timing of both
spring and fall migration in response to weather conditions, but

the responses vary between species and even populations in
terms of which weather variables are associated with migration
and to what extent. For example, warmer temperatures are
associated with earlier arrival on the breeding grounds in
Ruby-throated Hummingbirds (Archilochus colubris (Linnaeus,
1758)) (Courter et al. 2013). In Yellow Warblers (Setophaga petechia
(Linnaeus, 1766)) (Drake et al. 2014), strong westerly winds appear to
slow migration and result in a later clutch initiation date, whereas
in the Yellow-breasted Chat (auricollis) (Icteria virens auricollis (Deppe,
1830)) (Huang et al. 2017), westerly winds are linked to a decline in
survival during migration and a later arrival date at the breeding
grounds.
The challenge in studying large-scale changes in migratory timing lies in the accessibility of multiyear datasets that encompass
broad geographic regions. Citizen science can collect data on a
much greater spatial and temporal scale than can be obtained by
individual scientists, and in some cases, may provide access to
private lands that may otherwise be inaccessible (Dickinson et al.
2010). Well-known citizen science programs such as eBird have
proven invaluable in examining relationships between weather
and bird migration for many species over large spatial scales
(Hurlbert and Liang 2012; Arab et al. 2016; Newson et al. 2016). For
small migratory species that are challenging to track over large
spatial areas, such as the Vaux’s Swift (Chaetura vauxi (J.K. Townsend,
1839)), citizen-science-based approaches may be the most effective
approach for examining patterns of migration.
The Vaux’s Swift is a long-distance aerial insectivorous bird
with a declining population. These birds winter in Mexico and
Central America, migrate up the west coast, flying during the day
and roosting at night in chimneys or old-growth trees, and breed
along the west coast from California (USA) to as far north as
southern Yukon (Canada) (Fig. 1) (Schwitters et al. 2020). During
the breeding season, the swifts construct nests in semi-circles
composed of twigs and saliva located in hollow trees or chimneys.
Published by Canadian Science Publishing

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

58

Because these birds are aerial insectivores, changes to flying insect
populations may be an important factor leading to their declines
(Nebel et al. 2010; Nocera et al. 2012; Pomfret et al. 2012). In addition,
changes to insect emergence timing or availability during spring
and fall may influence their migration phenology. Vaux’s Swifts
roost with hundreds to thousands of other conspecifics in the cavities of hollowed-out old-growth trees and masonry chimneys during
migration (Shuford and Gardali 2008; Schwitters et al. 2020), and the
loss of old-growth trees may be another contributing factor to the
declining population of the species (Bull 2003). Given Vaux’s Swifts
reliance on roosts, a species-specific survey that targets roosts may
generate more comprehensive data than community-wide bird
surveys, such as eBird, for monitoring Vaux’s Swifts during
migration. Vaux’s Happening (https://www.vauxhappening.org/) has
been in operation since 2008 and involves volunteer monitors
counting Vaux’s Swifts at known roosts (typically decommissioned
chimneys) during spring and fall migration.
In this study, we made use for the first time of a citizen science
project specific to the Vaux’s Swift, Vaux’s Happening, to examine trends in migration over time and the relationship between
migratory timing and weather in this species. We predicted that
we would detect an advancement in spring migratory timing,
likely due to the earlier onset of spring events in the northern
hemisphere (Parmesan 2007). However, due to the pressure to
migrate quickly and arrive early to the breeding grounds (Alerstam
2011; Karlsson et al. 2012; Nilsson et al. 2013), we predicted that
Vaux’s Swifts would be relatively unresponsive to weather conditions experienced en route to the breeding grounds. In contrast to
spring migration, we predicted weaker, if any, changes in fall migratory timing across the study period (Brisson-Curadeau et al.
2019). However, we predicted that weather conditions experienced
en route would be more strongly correlated with fall migratory
timing due to comparatively reduced pressure to arrive early to the
wintering grounds, which would result in birds adjusting their
migratory timing based on local weather conditions.

Materials and methods
Migration data
Roost counts of Vaux’s Swifts during spring and fall migration
were collected by the citizen science organization Vaux’s Happening, a coordinated effort involving 350 citizen scientists (https://
www.vauxhappening.org/). Volunteer monitors visited sites in which
Vaux’s Swifts have historically roosted (Fig. 1) and recorded the
number of swifts roosting each day during northward and
southward migration, with counts ranging from a few swifts to
tens of thousands. To ensure that first arrival dates were recorded,
monitors began visiting roosts several days before birds were
projected to arrive at the site. Whenever possible, eBird was
used to track where swifts had already been sighted and project
when they would arrive at a given roost. Once the first swifts
were sighted, monitors began nightly roost checks (when possible).
In good weather, the mean flock entry typically began 2 min after
sunset. Monitors were instructed to arrive 30 min before sunset and
to stay at the roost for 30 min after sunset if no swifts arrived.
Experienced observers were paired with new observers for their
first-time monitoring roosts. Only swifts entering roosts were
counted. When large numbers of swifts were observed, monitors
either counted by tens using a clicker or counts were estimated
at a rate of 10 swifts/s entering the roost. This rate was calculated
from two live camera streams at the Monroe Wagner, Washington,
USA, roost. At this site, counts from the camera were also used
during poor weather and low visibility, by estimating the number of
swifts exiting the roost. In this case, the count of swifts exiting the
roost was estimated by multiplying the length of video time (in
seconds) during which swifts were exiting by 8 (as above, slow-

Can. J. Zool. Vol. 100, 2022

motion video was used to estimate an exit rate of 8 birds/s). When
camera counts and counts by monitors on site were compared, the
results were similar. At the end of the season, data were submitted
to the Vaux’s Happening database.
Data cleaning and segmentation
We analyzed spring migration data ranging from 2008 to 2017
and fall migration data from 2008 to 2016. From these data, we
extracted information on first arrival date (the date on which the
first swift arrived at the roost) and peak roost occupancy (the
date on which the maximum number of swifts were observed at
that roost during the season). First arrival dates may represent an
individual’s response to environmental conditions and not the
population as a whole, and the reliability of the data is sensitive
to population size and observer effort (Miller-Rushing et al.
2008); therefore, we also included peak roost occupancy as an indication of a population wide respond to environmental conditions. Spring and fall seasons were delineated each year based on
known species life history (Schwitters et al. 2020) and through
examining the distribution of count data at northern roosts,
which indicated a strong bimodal distribution with a gap occurring in mid-July (for example of data distribution see Supplementary Fig. S1),1 although exact dates for spring and fall migration
varied from year to year. Furthermore, the start of fall migration
was confirmed each year by examining camera footage from the
Monroe Wagner roost.
We restricted our analysis to roosts that had at least five observations during a season in at least 3 years, which resulted in
155 first arrival and peak roost occupancy observations in the
spring, at 26 different roosts, and 148 first arrival and peak roost
occupancy observations in the fall, at 28 different roosts. Because
data collection was variable across sites (e.g., some sites are remote
and difficult to access, whereas others, such as Chapman Elementary in Portland, Oregon, USA, attract dozens of spectators nightly
during migration), we decided on an approach that would allow us
to look at change over time (at least 3 years of observations) and
assess peak roost occupancy (at least five observations within a
season), while capturing a broad array of roosts with varying
numbers of swifts rather than limit our data to only accessible
sites with large numbers of swifts. Though this approach may
add a degree of uncertainty to our analysis, it is unlikely to
introduce systematic bias (i.e., systematically recording earlier
or later arrival dates). Overall, during spring migration, monitors performed, on average, 467 counts per year, with a minimum of 224 counts and a maximum of 702 counts. During fall
migration, monitors performed, on average, 367 counts per
year, with a minimum of 178 counts and a maximum of 506
counts.
Weather data
We obtained data on weather conditions at both a local and a
regional scale. For local weather, mean temperature (°C), precipitation (mm), wind speed (m/s), and wind gust speed (m/s) were calculated for the 3 weeks prior to the mean peak roost occupancy
date at each roost. Weather data were retrieved from Weather
Underground (https://www.wunderground.com/) and recorded
from the nearest airport to the roost. Weather Underground
recorded historical weather data from local airports taken in real
time at 1, 3, or 6 h intervals, which varies depending on each
weather station. We did not include historical weather data from
airports that were greater than 100 km away from the roost
because it may not have been an accurate representation of the
weather experienced at the roost. Unfortunately, wind direction
was not available from each airport; therefore, we were not able
to include wind direction at a local scale. However, we calculated
wind at the regional scale (en route to the roost) to test the

Supplementary figures are available with the article at https://doi.org/10.1139/cjz-2021-0089.

1

Published by Canadian Science Publishing

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

Prytula et al.

influence of southerly (Vwind) and westerly (Uwind) wind speed
(m/s) from a larger area on first arrival and peak roost occupancy
during spring and fall migration, following Drake et al. (2014).
Using the RNCEP program (Kemp et al. 2012), regional wind data
were downloaded from the National Center for Environmental
Prediction (NCEP) and were averaged over 2.5° latitude by longitude
cells across North America, with a temporal resolution of 6 h. We
chose to use a 3-week window to reflect mean conditions
experienced prior to peak roost occupancy for several reasons.
For Vaux’s Swifts, we do not have data on the migration speed
of individuals and some individuals may spend several days or
longer at individual roosts along the migratory path. In addition,
our goal was to capture the conditions that the birds would
experience as they migrated towards the roost, and as such, we
chose to average over a longer period of time rather than obtain
a simple snapshot that is subject to high variability in conditions.
Finally, because Vaux’s Swifts are aerial insectivores, conditions
experienced in the 3 weeks preceding arrival to the roost may
affect insect abundance and availability — factors that are key to
fueling during migration. To capture wind conditions experienced
en route during spring migration, spring NCEP wind data were
averaged over an area of 15° of latitude and 10° of longitude south
of each roost (Supplementary Fig. S2).1 To capture wind conditions
experienced en route during fall migration, NCEP wind data were
averaged over an area of 10° of latitude and 10° of longitude north
of each roost (Supplementary Fig. S2).1 For spring migration, 15° of
latitude ensured that we did not collect wind data from the
wintering grounds. For fall migration, 10° of latitude ensured that
we did not exceed the northern bound of the breeding range. Thus,
weather data included the following variables: local wind, wind
gust, precipitation, temperature, and regional southerly (Vwind)
and westerly (Uwind) wind.
Statistical analysis
To examine potential changes over time and associations
between weather and timing of first arrival and peak roost occupancy, we constructed a series of linear mixed models (LMMs).
Models were constructed so that each variable was tested on its
own and with various combinations of other variables. The full
model for either spring or fall included local wind (Wind; m/s),
local wind gust (WindGust; m/s), local precipitation (Precip; mm),
local temperature (Temp; °C), regional southerly wind (Vwind;
m/s), regional westerly wind (Uwind; m/s), latitude (Lat), and year
(Year). The models that we evaluated included various combinations of the above variables to evaluate all combinations. This
resulted in a total of 28 models for each season, plus a null model
with the intercept only. Roost was considered a random effect in
each model.
All of the considered effects were additive and did not include
interactive effects. We tested for multicollinearity among fixed
effects, but there were no strong correlations between variables
(r < 0.56), and variance inflation factors of fixed effects were
all <2. We used Akaike’s information criterion corrected for
small sample size (AICc) to rank the models, and we present
model-averaged 95% confidence intervals (95% CI) for models
within 4 DAICc of the top model (Burnham and Anderson 2002).
The statistical software R version 4.0.2 (R Core Team 2021) was
used to analyze the data, including the packages AICcmodavg
 2009),
(Mazerolle 2020), lme4 (Bates et al. 2015), MuMIn (Barton
and MASS (Venables and Ripley 2002).

Results
Spring migration
First arrival
The top model explaining variation in first arrival during
spring migration included effects of year, local wind gust, and latitude (Table 1). Other models within 4 DAICc units included the

59

Table 1. Summary of top-ranked models (<4 DAICc from top-ranked
model), using Akaike’s information criterion corrected for small sample
size (AICc), that explain the variation in migratory timing of Vaux’s
Swifts (Chaetura vauxi) during spring migration, 2008–2017.
AICc

DAICc

AICc weight

First arrival
Year + WindGust + Lat
Year + Wind + WindGust + Lat
Year + Precip + WindGust + Lat
Year + Wind + WindGust + Vwind + Lat
Year + Precip + Wind + WindGust + Lat

852.91
854.03
855.17
855.84
856.32

0
1.12
2.26
2.93
3.41

0.33
0.19
0.11
0.08
0.06

Peak roost occupancy
Year + Precip + Temp + Lat
Year + Precip + Lat
Year + Lat
Year + Precip + Temp + WindGust + Lat
Lat
Year + WindGust + Lat
Year + Precip + Temp + Uwind + Lat

863.06
863.45
863.45
863.61
863.79
864.03
864.06

0
0.39
0.4
0.56
0.74
0.98
1.01

0.1
0.08
0.08
0.08
0.07
0.06
0.06

Note: Roost is included in all models as a random effect. The AICc of the
model, difference between the model and the top model’s AICc (DAICc), and the
weight of each model (AICc weight) are shown. Vwind is regional southerly
wind (m/s); Uwind is regional westerly wind (m/s); Precip is local precipitation
(mm); Temp is local temperature (°C); Lat is latitude; Wind is local wind (m/s);
WindGust is local wind gust (m/s).

effects of local wind, regional southerly wind, and precipitation
in addition to those effects found in the top model. However,
only the 95% CI for the effects of year, wind gust, and latitude did
not overlap zero (Table 2). Swifts arrived 0.9 days earlier per year
and 0.37 days earlier per unit increase (m/s) in wind gust. Swifts
arrived 1.33 days later per 1° increase in latitude.
Peak roost occupancy
The top model explaining variation in peak roost occupancy
during spring migration included effects of year, precipitation,
temperature, and latitude (Table 1). Other models within 4 DAICc
units included the effects local wind and wind gust, and regional
westerly and southerly wind in addition to those effects found in
the top model. The 95% CI for the effects of year and latitude did
not overlap zero (Table 2) and indicated that swifts arrived
0.77 days earlier per year and 1.07 days later per 1° increase in
latitude.
Fall migration
First arrival
The top model explaining variation in first arrival during fall
migration included effects of temperature, regional southerly
and westerly wind, and local wind (Table 3). Other models within
4 DAICc units included the effects of year, temperature, precipitation, and local wind gust in addition to those effects found in the
top model. The 95% CI for the effects of temperature, local wind,
and latitude did not overlap zero (Table 2) and indicated that
swifts arrived 1.7 days later per 1° increase in temperature,
1.33 days earlier per unit increase (m/s) in local wind speed, and
1.07 days earlier per 1° increase in latitude.
Peak roost occupancy
The top model explaining variation in peak roost occupancy
during fall migration included effects of temperature, regional
westerly and southerly wind, and local wind (Table 3). Other models within 4 DAICc units included the effects of year, temperature,
precipitation, and local wind gust in addition to those effects
found in the top model. The 95% CI for the effects of temperature
Published by Canadian Science Publishing

60

Can. J. Zool. Vol. 100, 2022

Table 2. Model-averaged parameter estimates and 95% confidence intervals (in parentheses) for variables included
in the top-ranked models (<4 DAICc units of best model) that explain the variation in Vaux’s Swift (Chaetura vauxi)
first arrival and peak roost occupancy for both spring and fall migration.
Spring migration

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

Year
Precip
Temp
Uwind
Vwind
Wind
WindGust
Lat

Fall migration

First arrival

Peak roost occupancy

First arrival

Peak roost occupancy

–0.90 (–1.58, –0.23)
0.043 (–1.14, 1.23)

–0.77 (–1.52, –0.01)
–0.85 (–2.12, 0.42)
0.61 (–0.29, 1.5)
–1.85 (–4.56, 0.87)
–0.36 (–2.59, 1.88)
–0.29 (–0.98, 0.39)
–0.14 (–0.4, 0.13)
1.07 (0.45, 1.70)

–0.76 (–1.87, 0.34)
0.62 (–1.92, 3.16)
1.7 (0.58, 2.82)
1.02 (–1.8, 3.83)
–0.95 (–3.16, 1.26)
–1.33 (–2.62, –0.04)
–0.085 (–0.54, 0.38)
–1.07 (–2.09, –0.05)

0.068 (–0.74, 0.88)
1.5 (–0.41, 3.41)
0.88 (0.09, 1.67)
–1.83 (–3.75, 0.086)
0.25 (–1.43, 1.93)
–1.23 (–2.11, –0.36)
–0.029 (–0.37, 0.31)
–0.87 (–1.75, 0.016)

–0.75 (–2.93, 1.43)
0.40 (–0.30, 1.10)
–0.37 (–0.64, –0.11)
1.33 (0.73, 1.94)

Note: Values in boldface type represent confidence intervals that do not overlap zero. Vwind is regional southerly wind (m/s);
Uwind is regional westerly wind (m/s); Precip is local precipitation (mm); Temp is local temperature (°C); Lat is latitude; Wind is
local wind (m/s); WindGust is local wind gust (m/s).

Table 3. Summary of top-ranked models (<4 DAICc from top-ranked model), using Akaike’s
information criterion corrected for small sample size (AICc), that explain the variation in
migratory timing of Vaux’s Swifts (Chaetura vauxi) during fall migration, 2008–2016.
AICc

DAICc

AICc weight

First arrival
Temp + Uwind + Vwind + Wind
Year + Precip + Lat
Temp
Year + Precip + Temp + Uwind + Lat
Temp + Uwind + Vwind + Wind
Temp + Uwind + Vwind + Wind + WindGust + Lat
Precip + Temp + Uwind + Vwind + Wind

683.15
683.75
684.09
684.47
685.53
685.56
685.57

0
0.6
0.94
1.32
2.38
2.41
2.42

0.13
0.1
0.08
0.07
0.04
0.04
0.04

Peak roost occupancy
Temp + Uwind + Vwind + Wind
Precip + Temp + Uwind + Vwind + Wind
Year + Precip + Temp + Wind + WindGust + Lat
Temp + Uwind + Vwind + Wind + WindGust
Year + Precip + Temp + Uwind + Vwind + Wind
Precip + Temp + Uwind + Vwind + Wind + WindGust
Precip + Temp + Uwind + Vwind + Wind + WindGust + Lat

626.01
626.77
628.33
628.36
629.24
629.24
629.56

0
0.75
2.31
2.35
3.22
3.23
3.54

0.2
0.14
0.06
0.06
0.04
0.04
0.03

Note: Roost is included in all models as a random effect. The AICc of the model, difference between the
model and the top model’s AICc (DAICc), and the weight of each model (AICc weight) are shown. Vwind is
southerly wind (m/s); Uwind is westerly wind (m/s); Precip is precipitation (mm); Temp is temperature (°C);
Lat is latitude; Wind is local wind (m/s); WindGust is local wind gust (m/s).

and local wind did not overlap zero (Table 2) and indicated that
swifts arrived 0.88 days later per 1° increase in temperature and
1.23 days earlier per unit increase (m/s) in local wind speed. It is
worth noting that although 95% CI for latitude and westerly wind
did not overlap zero, they were close to doing so (<0.1 overlap of
zero). Swifts tended to arrive earlier at higher latitudes and with
stronger westerly winds.

Discussion
In this study, we explored a citizen science dataset spanning
10 years to examine potential changes in migration over time
and associations between weather (at both a local and a regional
scale) and the timing of arrival at migratory roosts in Vaux’s
Swifts. We found support for our prediction that Vaux’s Swift arrival date at migratory roosts advanced significantly during
spring migration from 2008 to 2017, but it has not changed during fall migration (from 2008 to 2016). Though there was little
association between weather conditions experienced en route
and migratory timing for spring migration (only wind gust had
was significant, and only for first arrivals), fall migration timing

was associated with both temperature and local wind conditions
experienced en route, for both first arrival and peak roost occupancy timing. Even over a relatively short time frame, Vaux’s Swifts
arrived at migratory roosts earlier in spring (both first arrival and
peak roost occupancy), which could be indicative of a response to
the advancement of spring phenology. This rapid change is not surprising because many species across a broad range of avian taxa
have exhibited significant advances in spring migratory timing,
likely in response to changing spring phenology (Coppack and
Both 2002; Both et al. 2005; Gordo and Sanz 2006; Mckinney et al.
2012).
Over a 10-year period, we found that Vaux’s Swifts migrated
earlier in spring, as evidenced by advancement in both peak
roost occupancy and first arrival dates at roosts. Vaux’s Swifts
may have responded also to local wind gust conditions, as they
showed earlier first arrivals when there were stronger wind gusts,
potentially because favourable tailwinds help to accelerate migration (Haest et al. 2019). However, first arrival dates can be biased
since they are influenced by individual behaviour, population size,
and sampling effort, and thus may not represent population
processes as a whole (Miller-Rushing et al. 2008). Furthermore,
Published by Canadian Science Publishing

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

Prytula et al.

as predicted, neither local temperature nor precipitation were
strong indicators of first arrival or peak roost occupancy timing
during spring migration. Together, these results are consistent
with findings in Chimney Swifts (Chaetura pelagica (Linnaeus,
1758)) and other long-distance migrants that have shown no
response to weather experienced during migration, and specifically, no response to changes in temperature (Zaifman et al.
2017).
We found stronger associations between migratory timing and
fall weather conditions experienced en route, which could suggest that Vaux’s Swifts are altering migratory timing and time
spent at migratory roosts to optimize survival during southward
migration (Jenni and Kery 2003). Unlike spring arrival, associations between weather and migratory timing were observed in
both first arrivals and peak roost occupancy, suggesting that they
may have been true population-level responses. Earlier fall migration dates were associated with lower temperatures, potentially
due to cold snaps resulting in reduced insect availability (Jenni
and Kery 2003). Higher local wind speeds measured at the closest
weather stations to roosts were also associated with earlier arrivals at migratory roosts. As above, local wind could affect migratory timing through favourable tailwinds (Haest et al. 2019),
although we were unable to measure the direction of local winds.
In addition, wind speed could affect stopover by decreasing flying
insect availability or capture efficiency (Møller 2013). Vaux’s Swift
peak roost occupancy also tended to be earlier with increased regional westerly winds, although the 95% CI for this effect overlapped zero and so any presumed effect is likely weak. Local wind
seemed to have a stronger association with migration timing than
regional wind patterns, which could be due to the large-scale
effects being less representative of what individual birds experience during migration compared with local weather. However,
Drake et al. (2014) found that high regional westerly winds at
stopover sites were associated with lower survival and later arrival to
the breeding grounds in Yellow Warblers, potentially due to an
association between westerly winds and extreme weather events
in Western Canada. Thus, the effects of weather events on migratory timing may be species- and season-specific and may be influenced by the migration ecology of the species (e.g., Vaux’s Swifts
propensity to use migratory roosts may help buffer against the
effects of extreme weather events).
Many studies of migratory timing in response to climate and
weather have focused on spring migration, while fewer have
focused on fall migration, likely because the complexity of drivers of fall migration and the prolonged nature of fall migration
make it challenging to study (Gallinat et al. 2015). Yet, understanding responses to climate change will require consideration
of avian phenology across the full annual cycle. Citizen science
programs such as eBird have become increasingly popular for
studying avian migration and the effects of climate change, as
they allow for the examination of migration timing over broad
geographic regions and long time periods (Arab et al. 2016;
La Sorte and Graham 2021). Programs such as Vaux’s Happening
that use a targeted approach in terms of counting a specific species
at designated locations annually likely allow for a greater and
more consistent detection of individuals and population-level
patterns for species such as Vaux’s Swifts that migrate to specific
known locations at high concentrations. However, since this program focuses on known historical roosts, most of which are artificial (i.e., decommissioned chimneys), our results may not extend
to other Vaux’s Swift populations that may be using unknown roosts
along the migratory route, such as old-growth trees in nonurban
areas. Furthermore, shifts in use between known and unknown
roosts could result in perceived changes in migration phenology
when the population as a whole may not be changing. Future studies
may benefit from combining Vaux’s Happening and eBird data to
overcome some of these issues.

61

We did not examine associations between weather conditions
experienced during the breeding or wintering period and migration timing, due to uncertainty in terms of where individual
swifts had bred or wintered. Migratory connectivity and the exact
wintering locations of Vaux’s Swifts remain poorly studied (Reudink
et al. 2015; Schwitters et al. 2020) and would benefit from additional
research using individual tracking techniques such as light-level
geolocators, banding, or stable isotope analysis. Individual tracking
would have had additional benefits in our study: specifically, without being able to identify individual swifts, it is possible that some
degree of temporal and spatial autocorrelation may have occurred
which we were not able to control for if the same swifts were using
multiple roosts across the broad spatial scales of our study area.
Follow-up studies should also examine the role of carry-over
effects from wintering or breeding ground conditions on migratory timing.
Many long-distance migrants initiate spring migration primarily
based on endogenous rhythms rather than response to environmental conditions (Both and Visser 2001), a characteristic that may make
them especially vulnerable to climate change and shifting spring
phenology. Our results suggest that Vaux’s Swifts may be adjusting
aspects of their migration based on external factors, which may
allow them to better adjust to rapidly changing spring phenology.
In addition, as aerial insectivores, swifts are vulnerable to changes
in insect populations and human use of pesticides (Nocera et al.
2012; Pomfret et al. 2012; Nebel et al. 2010). Finally, Vaux’s Swifts
rely on old-growth trees and chimneys for roosting or nesting, both
of which are declining due to continued deforestation and changing construction techniques as well as the decommissioning of old
chimneys (Bull 2003).
In conclusion, we documented a rapid advance in Vaux’s Swifts
spring migration timing and stronger associations between migratory timing and weather experienced en route during fall
migration compared with spring migration. Questions remain as
to the causal factors resulting in advanced spring migration, and
whether the changes are a result of phenotypic plasticity, or if
they are a long-term adaptation that will be able to keep up with
the changing climate and prevent further population decline
(Møller et al. 2008). The lack of relationship between migratory
timing and spring weather conditions experienced en route
could potentially be problematic since climate change could
result in more extreme weather conditions. An inability to
adjust phenology to extreme weather events could result in a
delay in spring arrival and survivorship during migration, much
like the cold spell that resulted in low rates of return to the
breeding grounds in the long-distance migrant Semicollared Flycatchers (Ficedula semitorquata (Homeyer, 1885) (Briedis et al.
2017). In contrast, we found strong relationships between migratory timing and weather conditions experienced en route during fall migration, which could allow Vaux’s Swifts to track
future changes in local weather patterns. With inclusion of the
effect of weather variables on wintering and breeding grounds,
we may learn more about what causes swifts to leave these wintering or breeding grounds, and not just what pushes or pulls
them among roosts.

Funding statement
Funding was provided by a Natural Sciences and Engineering
Research Council of Canada (NSERC) Discovery Grant to M.W.R., and
an Irving K. Barber BC Indigenous Student Award, New Relationship Trust Scholarship, Indspire, and the Ken Lepin Award to E.P.

Acknowledgements
We thank E. Smith, S. Pierera, and D. Agar for their help processing
data, and J. Dunne-Mucklow for GIS support and map creation.
Special thanks go to A. Drake for assistance with collecting and
analyzing regional wind data. We also thank the many volunteers
Published by Canadian Science Publishing

62

who assisted with data collection for Vaux’s Happening over the
years. Research was conducted at Thompson Rivers University and
took place on the traditional lands of the Tk’emlúps te Secwépemc
within Secwépemc’ulucw, the traditional and unceded territory of
the Secwépemc people.

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

References
Alerstam, T. 2011. Optimal bird migration revisited. J. Ornithol. 152: 5–23.
doi:10.1007/s10336-011-0694-1.
Arab, A., Courter, J.R., and Zelt, J. 2016. A spatio-temporal comparison of
avian migration phenology using citizen science data. Spatial Stat. 18:
234–245. doi:10.1016/j.spasta.2016.06.006.
 , K. 2009. Mu-MIn: multi-model inference. R package version 0.12.2/r18.
Barton
Available from https://R-Forge.R-project.org/projects/mumin/.
Bates, D., Mächler, M., Bolker, B., and Walker, S. 2015. Fitting linear mixedeffects models using lme4. J. Stat. Softw. 67(1): 1–48. doi:10.18637/jss.v067.i01.
Bearhop, S., Fiedler, W., Furness, R.W., Votier, S.C., Waldron, S., Newton, J.,
et al. 2005. Assortative mating as a mechanism for rapid evolution of a
migratory divide. Science, 310: 502–504. doi:10.1126/science.1115661. PMID:
16239479.
BirdLife International and Handbook of Birds of the World. 2019. Bird species distribution maps of the World. Version 2019.1. Available from http://
datazone.birdlife.org/species/requestdis.
Both, C., and Visser, M.E. 2001. Adjustment to climate change is constrained
by arrival date in a long-distance migrant bird. Nature, 411: 296–298. doi:10.
1038/35077063. PMID:11357129.
Both, C., Bijlsma, R.G., and Visser, M.E. 2005. Climatic effects on spring
migration and breeding in a long-distance migrant, the pied flycatcher
Ficedula hypoleuca. J. Avian. Biol. 36: 368–373. doi:10.1111/j.0908-8857.2005.
03484.x.
Bozó, L., Csörgo†, T., and Heim, W. 2018. Weather conditions affect spring
and autumn migration of Siberian leaf warblers. Avian Res. 9: 33. doi:10.1186/
s40657-018-0126-5.
Bridge, E.S., Ross, J.D., Contina, A.J., and Kelly, J.F. 2015. Do molt–migrant
songbirds optimize migration routes based on primary productivity? Behav.
Ecol. 27: 784–789. doi:10.1093/beheco/arv199.
Briedis, M., Hahn, S., and Adamik, P. 2017. Cold spell en route delays spring
arrival and decreases apparent survival in a long-distance migratory
songbird. BMC Ecol. 17: 11. doi:10.1186/s12898-017-0121-4. PMID:28376915.
 Elliott, K.H., and Côté, P. 2019. Factors influencing fall
Brisson-Curadeau, E.,
departure phenology in migratory birds that bred in northeastern North
America. Auk, 137(1): ukz064. doi:10.1093/auk/ukz064.
Brown, C.R., and Brown, M.B. 1998. Intense natural selection on body size
and wing and tail asymmetry in cliff swallows during severe weather.
Evolution, 52: 1461–1475. doi:10.1111/j.1558-5646.1998.tb02027.x. PMID:28565386.
Bull, E.L. 2003. Declines in the breeding population of Vaux’s Swifts in
Northeastern Oregon. Western Birds, 34: 230–234.
Burnham, K.P., and Anderson, D.R. 2002. Model selection and multimodel
inference: a practical information-theoretic approach. 2nd ed. SpringerVerlag, New York.
Cook, J., Nuccitelli, D., Green, S.A., Richardson, M., Winkler, B., Painting, R.,
et al. 2013. Quantifying the consensus on anthropogenic global warming
in the scientific literature. Environ. Res. Lett. 8: 024024. doi:10.1088/17489326/8/2/024024.
Coppack, T., and Both, C. 2002. Predicting life-cycle adaptation of migratory
birds to global climate change. In The avian calendar: exploring biological hurdles in the annual cycle. Proc. 3rd Conf. European Om. Union, Groningen, August 2001. Edited by C. Both and T. Piersma. Ardea, 90: 369–378.
doi:10.5253/arde.v90i3.p369.
Covino, K.M., Horton, K.G., and Morris, S.R. 2020. Seasonally specific changes in
migration phenology across 50 years in the Black-throated Blue Warbler. Auk,
137: 1–11. doi:10.1093/auk/ukz080.
Courter, J.L., Johnson, R.J., Bridges, W.C., and Hubbard, K.G. 2013. Assessing
migration of Ruby-Throated Hummingbirds (Archilochus colubris) at broad
spatial and temporal scales. Auk, 130: 107–117. doi:10.1525/auk.2012.12058.
Cristol, D.A. 1995. Early arrival, initiation of nesting, and social status: an
experimental study of breeding female red-winged blackbirds. Behav. Ecol. 6:
87–93. doi:10.1093/beheco/6.1.87.
Dickinson, J.L., Zuckerberg, B., and Bonter, D.N. 2010. Citizen science as an
ecological research tool: Challenges and benefits. Annu. Rev. Ecol. Evol.
Syst. 41: 149–172. doi:10.1146/annurev-ecolsys-102209-144636.
Drake, A., Rock, C.A., Quinlan, S.P., Martin, M., and Green, D.J. 2014. Wind
speed during migration influences the survival, timing of breeding, and
productivity of neotropical migrant, Setophaga petechiai. PLoS ONE, 9:
e97152. doi:10.1371/journal.pone.0097152. PMID:24828427.
Gallinat, A.S., Primack, R.B., and Wagner, D.L. 2015. Autumn, the neglected
season in climate change research. Trends Ecol. Evol. 30: 169–176. doi:10.1016/j.
tree.2015.01.004.
Gordo, O., and Sanz, J.J. 2006. Climate change and bird phenology: a longterm study in the Iberian Peninsula. Global Change Biol. 12: 1993–2004.
doi:10.1111/j.1365-2486.2006.01178.x.

Can. J. Zool. Vol. 100, 2022

Haest, B., Huppop, O., Pol, M., and Bairlein, F. 2019. Autumn bird migration
phenology: A potpourri of wind, precipitation and temperature effects.
Global Change Biol. 25: 4064–4080. doi:10.1111/gcb.14746.
Huang, A.C., Bishop, C.A., McKibbin, R., Drake, A., and Green, D.J. 2017.
Wind conditions on migration influence the annual survival of a neotropical migrant, the western yellow-breasted chat (Icteria virens auricollis).
BMC Ecol. 17: 29. doi:10.1186/s12898-017-0139-7.
Hurlbert, A.H., and Liang, Z.A. 2012. Spatiotemporal variation in avian migration
phenology: Citizen science reveals effects of climate change. PLoS ONE, 7:
e31662. doi:10.1371/journal.pone.0031662. PMID:22384050.
Jenni, L., and Kery, M. 2003. Timing of autumn bird migration under climate
change: advances in long-distance migrants, delays in short-distance migrants.
Proc. R Soc. B Biol. Sci. 270: 1467–1471. doi:10.1098/rspb.2003.2394.
Karlsson, H., Nilsson, C., Backman, J., and Alerstam, T. 2012. Nocturnal
passerine migrants fly faster in spring than in autumn: a test of the time
minimization hypothesis. Anim. Behav. 83: 87–93. doi:10.1016/j.anbehav.2011.
10.009.
Kelly, J.F., Horton, K.G., Stepanian, P.M., de Beurs, K.M., Fagin, T., Bridge, E.S.,
and Chilson, P.B. 2016. Novel measures of continental-scale avian migration
phenology related to proximate environmental cues. Ecosphere, 7: e01434.
doi:10.1002/ecs2.1434.
Kemp, M.U., van Emiel, L.E., Shamoun-Baranes, J., and Bouten, W. 2012.
RNCEP: Global weather and climate data at your fingertips. Methods Ecol.
Evol. 3: 65–70. doi:10.1111/j.2041-210X.2011.00138.x.
La Sorte, F.A., and Graham, C.H. 2021. Phenological synchronization of seasonal bird migration with vegetation greenness across dietary guilds. J. Anim.
Ecol. 90: 343–355. doi:10.1111/1365-2656.13345. PMID:33107060.
Mazerolle, M.J. 2020. AICcmodavg: Model selection and multimodel inference based on (Q)AIC(c). R package version 2.3-1. Available from https://
CRAN.R-project.org/package=AICcmodavg.
McKinney, A.M., CaraDonna, P.J., Inouye, D.W., Barr, B., Bertelsen, C.D., and
Waser, N.M. 2012. Asynchronous changes in phenology of migrating Broadtailed Hummingbirds and their early-season nectar resources. Ecology, 93:
1987–1993. doi:10.1890/12-0255.1. PMID:23094369.
Miller-Rushing, A., Lloyd-Evans, T., Primack, R., and Satzinger, P. 2008. Bird
migration times, climate change, and changing population sizes. Global
Change Biol. 14(9): 1959–1972. doi:10.1111/j.1365-2486.2008.01619.x.
Møller, A.P. 2013. Long-term trends in wind speed, insect abundance and
ecology of an insectivorous bird. Ecosphere, 4: 1–11. doi:10.1890/ES12-00310.1.
Møller, A.P., Rubolini, D., and Lehikoinen, E. 2008. Populations of migratory
bird species that did not show a phenological response to climate change
are degrading. Proc. Nat. Sci. U.S.A. 105: 16195–16200. doi:10.1073/pnas.
0803825105.
Nathan, R., Getz, W.M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D., and
Smouse, P.E. 2008. A movement ecology paradigm for unifying organismal movement research. Proc. Natl. Acad. Sci. U.S.A. 105: 19052–19059.
doi:10.1073/pnas.0800375105.
Nebel, S., Mills, A., McCracken, J.D., and Taylor, P.D. 2010. Declines of aerial
insectivores in North America follow a geographic gradient. Avian Conserv.
Ecol. 5(2): 1. doi:10.5751/ACE-00391-050201.
Newson, S.E., Moran, N.J., Musgrove, A.J., Pearce-Higgins, J.W., Gillings, S.,
Atkinson, P.W., et al. 2016. Long-term changes in the migration phenology
of UK breeding birds detected by large-scale citizen science recording
schemes. Ibis, 158: 481–495. doi:10.1111/ibi.12367.
Nilsson, C., Klaassen, R.H., and Alerstam, T. 2013. Differences in speed and
duration of bird migration between spring and autumn. Am. Nat. 181:
837–845. doi:10.1086/670335. PMID:23669545.
Nocera, J.J., Blais, J.M., Beresford, D.V., Finity, L.K., Grooms, C., Kimpe, L.E.,
et al. 2012. Historical pesticide applications coincided with an altered
diet of aerially foraging insectivorous chimney swifts. Proc. R Soc. B Biol.
Sci. 279: 3114–3120. doi:10.1098/rspb.2012.0445. PMID:22513860.
Parmesan, C. 2007. Influences of species, latitudes and methodologies on
estimates of phenological response to global warming. Global Change
Biol. 13: 1860–1872. doi:10.1111/j.1365-2486.2007.01404.x.
Pomfret, J.K., Nocera, J.J., Kyser, T.K., and Reudink, W.M. 2012. Linking population declines with diet quality in Vaux’s swifts. Northwest Sci. 88:
305–313. doi:10.3955/046.088.0405.
R Core Team. 2021. R: a language and environment for statistical computing.
R Foundation for Statistical Computing, Vienna, Austria. Available from
https://www.R-project.org/.
Reudink, M.W., Van Wilgenburg, S.L., Steele, L.S., Pillar, A.G., Marra, P.P.,
and McKellar, A.E. 2015. Patterns of migratory connectivity in Vaux’s Swifts at
a northern migratory roost: a multi-isotope approach. Condor, 117: 670–682.
doi:10.1650/CONDOR-15-82.1.
Rubolini, D., Spina, F., and Saino, N. 2004. Protandry and sexual dimorphism in trans-Saharan migratory birds. Behav. Ecol. 15: 592–601. doi:10.1093/
beheco/arh048.
Schwartz, M.D., Ahas, R., and Aasa, A. 2006. Onset of spring starting earlier
across the northern hemisphere. Global Change Biol. 12: 343–351. doi:10.1111/
j.1365-2486.2005.01097.x.
Schwitters, L., Schwitters, D., Bull, E.L., and Collins, C.T. 2020. Vaux’s Swift
(Chaetura vauxi), version 1.0. In Birds of the world. Edited by P.G. Rodewald.
Cornell Lab of Ornithology, Ithaca, N.Y. doi:10.2173/bow.vauswi.01.
Published by Canadian Science Publishing

Prytula et al.

Venables, W.N., and Ripley, B.D. 2002. Modern applied statistics with S.
4th ed. Springer, New York. Available from https://www.stats.ox.ac.uk/
pub/MASS4/.
Zaifman, J., Shan, D., Ay, A., and Jimenez, A.G. 2017. Shifts in bird migration
timing in North American long-distance migrants are associated with climate change. Int. J. Zool. 2017: 1–9. doi:10.1155/2017/6025646.

Can. J. Zool. Downloaded from cdnsciencepub.com by CSP Customer Service on 12/07/21
For personal use only.

Shuford, W.D., and Gardali, T. 2008. California bird species of special concern: A
ranked assessment of species, subspecies, and distinct populations of birds of
immediate conservation concern in California. Stud. West. Birds, 1: 254–259.
Si, Y., Xin, Q., de Boer, W.F., Gong, P., Ydenberg, R.C., and Prins, H.H.T.
2015. Do arctic breeding geese track or overtake a green wave during
spring migration? Sci. Rep. 5: 8749. doi:10.1038/srep08749. PMID:25735996.

63

Published by Canadian Science Publishing