Content uploaded by Martins Briedis
Author content
All content in this area was uploaded by Martins Briedis on Mar 19, 2020
Content may be subject to copyright.
MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 638: 1–12, 2020
https://doi.org/10.3354/meps13274 Published March 19
1. INTRODUCTION
Bird migration usually comprises an interchange of
an active flying phase, when energy is consumed,
and a stationary stopover phase, when fuel is
restored by food intake (Alerstam & Lindström 1990,
Hedenström & Alerstam 1998). This is especially true
© The authors 2020. Open Access under Creative Commons by
Attribution Licence. Use, distribution and reproduction are un -
restricted. Authors and original publication must be credited.
Publisher: Inter-Research · www.int-res.com
*Corresponding author: martins.briedis@vogelwarte.ch
FEATURE ARTICLE
Seasonally specific responses to wind patterns and
ocean productivity facilitate the longest animal
migration on Earth
Tereza Hromádková1,2, Václav Pavel2, Jirˇí Flousek3, Martins Briedis4, 5, 6,*
1Department of Zoology, Faculty of Science, University of South Bohemia, 370 05 >eské Budeˇjovice, Czech Republic
2Centre for Polar Ecology, Faculty of Science, University of South Bohemia, 370 05 >eské Budeˇjovice, Czech Republic
3The Krkonoše Mountains National Park, Dobrovského 3, 543 01 Vrchlabí, Czech Republic
4Department of Zoology, Palacký University, 771 46 Olomouc, Czech Republic
5Swiss Ornithological Institute, 6204 Sempach, Switzerland
6Lab of Ornithology, Institute of Biology, University of Latvia, 2169 Salaspils, Latvia
ABSTRACT: Migratory strategies of animals are
broadly defined by species’ eco-evolutionary dynam-
ics, while behavioural plasticity according to the
immediate environmental conditions en route is cru-
cial for energy efficiency and survival. The Arctic tern
Sterna paradisaea is known for its remarkable migra-
tion capacity, as it performs the longest migration
known by any animal. Yet, little is known about the
ecology of this record-breaking journey. Here, we
tested how individual migration strategies of Arctic
terns are adapted to wind conditions and fuelling
opportunities along the way. To this end, we deployed
geolocators on adult birds at their breeding sites in
Svalbard, Norway. Our results confirm fundamental
predictions of optimal migration theory: Arctic terns
tailor their migration routes to profit from (1) tailwind
support during the movement phase and (2) food-rich
ocean areas during the stopover phase. We also found
evidence for seasonally distinct migration strategies:
terns prioritize fuelling in areas of high ocean produc-
tivity during the southbound autumn migration and
rapid movement relying on strong tailwind support
during the northbound spring migration. Travel speed
in spring was 1.5 times higher compared to autumn,
corresponding to an increase in experienced wind
support. Furthermore, with their pole-to-pole migra-
tion, Arctic terns experience approximately 80 % of all
annual daylight on Earth (the most by any animal),
easing their strictly diurnal foraging behaviour. How-
ever, our results indicate that during migration day-
light duration is not a limiting factor. These findings
provide strong evidence for the importance of interac-
tion between migrants and the environment in facili-
tating the longest animal migration on Earth.
O
PEN
PEN
A
CCESS
CCESS
Geolocator tagged Arctic term Sterna paradisaea at a breed-
ing site in Svalbard, Norway.
Photo: Martins Briedis
KEY WORDS: Arctic tern · Sterna paradisaea ·
Daylength · Geolocator · Global wind systems · Long-
distance migration · Migration phenology · Migration
strategy · Ocean productivity
Mar Ecol Prog Ser 638: 1–12, 2020
for flapping migrants (e.g. waders and songbirds)
that fuel their flight by burning energy, as opposed to
soaring migrants (e.g. raptors and albatrosses) that
often take advantage of thermals or topography-
induced air uplifts, to cover large distances with min-
imal energy expenditure. For flapping flyers, this
means that, while on migration, extended time needs
to be spent at stopover sites refuelling and storing
energy reserves to power the upcoming flight bouts
(Hedenström & Alerstam 1997). Because the rate of
energy expenditure during the flight phase is usually
much higher than the rate of fuel deposition on
stopovers, visiting high-quality stopover areas en
route is of high importance for efficient and success-
ful migration (Alerstam 2011). One of the main ways
to reduce the expenditure of stored energy during
flapping flight is to exploit prevailing wind regimes
and adjust migration routes to take a full advantage
of wind assistance for covering large distances rap-
idly (Kranstauber et al. 2015). Thus, maximizing
fuelling rates on stopovers and minimizing energy
expenditure during the flight bouts are the 2 key
aspects of migration energetics for species that use
powered flight for their migration (Alerstam 2003).
The overall migration duration of birds is largely de-
pendent on fuelling rates (Lindström 2003), while tail-
wind support can significantly facilitate faster travel-
ling during the movement phase (Green 2004).
Daylength regime (day:night ratio) has also been sug-
gested to have considerable effect on the total migra-
tion duration of diurnally foraging species because
of limited daylight hours available for fuelling
(Bauchinger & Klaassen 2005). As the season pro-
gresses and migrants move across latitudes, they ex-
perience changes in the daylength regime, which can
slow down or speed up their migration depending on
the time of the year, latitude and species’ activity
cycle (nocturnal vs. diurnal migrants and nocturnal
vs. diurnal foragers). For most species, migration
should be shorter (and faster) when daylength is
longer and there is more time for fuelling (Kvist &
Lindström 2000, Bauchinger & Klaassen 2005).
All these aspects should be especially important for
trans-equatorial migrants that cover long distances be-
tween their breeding and non-breeding areas. On the
extreme range of the migration distance spectrum is
the Arctic tern Sterna paradisaea that annually mi-
grates over 50 000 km between the breeding areas in
the Arctic and the non-breeding areas in Antarctica
(Storr 1958, Egevang et al. 2010). Despite a recent
surge in tracking studies on Arctic terns (Egevang et al.
2010, Fijn et al. 2013, McKnight et al. 2013, Volkov et
al. 2017, Alerstam et al. 2019), significant gaps remain
in our understanding about the ecology of this record-
breaking migration. To travel such vast distances
across all climatic zones, exploitation of environmental
settings for efficient travel may be crucial, as the seem-
ingly favourable shortest route may not be optimal. By
arching migration corridors according to the tailwind
support and further fine-tuning the routes to pass
through areas of high ocean productivity, migrants can
sustain the energy-demanding flight phase by stopping
over and refuelling at sites where food resources are
abundant. Furthermore, while residing at the breeding
and non-breeding sites, Arctic terns are exposed to
24 h daylight, enabling flexible interchange between
foraging and resting time. In contrast, during migration
terns are exposed to shifts between day and night as
they pass through intermediate latitudes. Because Arc-
tic terns are strictly diurnal foragers (McKnight et al.
2013), feeding and refuelling during migration are lim-
ited to the available daylight hours. This may force
terns to feed during the day and travel at night if rapid
and short-duration migration is advantageous.
In this study, we used geolocator tracking to, first,
describe migration patterns and non-breeding areas
of the Arctic terns breeding in the high Arctic (Sval-
bard, Norway) at the northern limits of the species’
distribution range (BirdLife International 2018). Sec-
ond, we tested how individual migration routes and
stopover sites are adapted to take advantage of wind
support and food availability en route. We hypothe-
sized that the longest animal migration on Earth is
facilitated by wind assistance during the flight phase
and abundance of food resources during the refu-
elling stopover phase, which is further eased by
extended daylength hours throughout the journey.
Hence, we predicted that (1) the chosen migration
routes of the tracked Arctic terns will be adapted so
that the birds benefit from tailwind support during
both southbound and northbound migrations; (2) the
stopover sites of the terns will be located in areas of
higher ocean productivity compared to passage
areas (McKnight et al. 2013); (3) the terns will time
their migration to cross the Equator near the
equinoxes, enabling longer foraging hours in both
hemispheres (Alerstam 2003).
2. MATERIALS AND METHODS
2.1. Field work
Our study site was located in Longyearbyen on
the island of Spitsbergen, Svalbard archipelago
(78° 14’ N, 15° 39’ E). We captured 30 breeding indi-
2
Hromádková et al.: Environmental effects on the longest animal migration
viduals during late stages of egg incubation between
8 and 14 July 2017, using tent spring traps placed on
their nests. All captured birds were marked with
unique colour ring combinations and equipped with
multi-sensor archival data loggers (geolocators;
model Intigeo-W65A9-SEA, Migrate Technology)
that were fixed to the colour rings. Geolocators were
set to sample ambient light intensity every minute
and store maximal values at 5 min intervals. Temper-
ature was sampled every 5 min, storing maximal and
minimal values at 4 h intervals; immersion and con-
ductivity measures were sampled every 30 s, storing
the sum of samples scored as wet and maximal con-
ductivity every 4 h.
The geolocators including colour rings weighed
1.06 ± 0.05 g (SD), which never exceeded 1.2% of the
body mass of the tagged birds (106.2 ± 7.6 g, n = 30).
Sex of all tagged individuals was determined molecu-
larly using a droplet of blood taken from the brachial
vein (Griffiths et al. 1998, Fridolfsson & Ellegren 1999,
Ležalová-Piálková 2011), as the accuracy of sexing
Arctic terns in the field is typically lower than 74%
(Fletcher & Hamer 2003, Devlin et al. 2004).
In the following season between 27 June and 12
July 2018, we managed to recapture 16 birds with
geolocators (53%) at the same breeding colony. At
least 7 more of the previously tagged birds were
sighted in the breeding colony (total return rate =
77%), but we failed to recapture them. This was
mainly due to high predation rate of the nests in this
season by Arctic fox Vulpes lagopus leading to fre-
quent relocation or disappearance of the individuals
whose nests were depredated.
Body mass of the tagged individuals was higher
upon recapture and removal of the geolocators com-
pared to the time of deployment a year earlier (2017:
104.6 ± 8 g [n = 16], mean capture date = 10 July;
2018: 110.9 ± 7.6 g, mean capture date = 4 July;
paired t-test: t14 = −2.691, p = 0.018; data met the
assumption of normality and homogeneity of vari-
ances based on a Shapiro-Wilk normality test and an
F-test, respectively). Thus, geolocators apparently
did not have a negative effect on the body condition
of the tagged individuals (Brlík et al. 2020).
2.2. Data analyses
2.2.1. Geolocator tracking
All data analyses were done in R version 3.5.1 (R
Core Team 2018). To calculate the geographic posi-
tions of the terns, we first log-transformed light inten-
sity recordings from the retrieved geolocators to
derive sunrise and sunset times (twilight events)
using the ‘preprocessLight’ function in the R-pack-
age ‘TwGeos’. Further, we used the R-package
‘FLightR’ to estimate geographic locations of the
tracked individuals (Rakhimberdiev et al. 2017). Dur-
ing both breeding and non-breeding periods,
tracked Arctic terns were exposed to 24 h daylight,
thus making it problematic to use these stationary
periods for calibrating the light data. Sunsets and
sunrises were recorded only during migration peri-
ods and during an approximately 2 mo long period
before the spring migration when the birds were at
an unknown location at the non-breeding sites.
Within the latter period, we identified extended peri-
ods when birds were stationary before the spring
migration by visually inspecting recorded sunrise
and sunset times. We then used the ‘find. stationary.
location’ function to estimate the geographic location
of this unknown site and used it for calibration.
Further, we followed standard analysis procedures
in ‘FLightR’ as outlined by Lisovski et al. (2020).
‘FLightR’ uses a template fit method to compute a
spatial likelihood surface for each twilight event. A
posterior distribution of the likeliest migration path
and its credible intervals are then derived via particle
filtering. Because the conductivity (salinity) record-
ing on all of our geolocators indicated that whenever
geolocators were immersed, birds were in saltwater,
we set 0 probability for birds occurring in areas that
were further than 50 km away from the shoreline. For
1 track (BH004, see Fig. S1e in the Supplement at
www. int-res. com/ articles/ suppl/ m638 p001 _ supp. pdf),
the imposed spatial mask led to a failure of the parti-
cle filter; therefore, the analyses for this geolocator
were run without a spatial mask. Finally, we used the
‘stationary.migration.summary’ function to deter-
mine stationary periods that were longer than 6 twi-
light events (3 d) and arrival/departure dates from
them. For this, we defined 20% as the minimum
probability of movement.
Using high-frequency data, McKnight et al. (2013)
showed that Arctic terns typically do not rest on
water, as only a small fraction of time (0.5 ± 0.22 h d−1
at most in August) is spent floating. Thus, immersion
recordings should approximately reflect the daily
rate of feeding dives. Based on this, we used the
cumulative daily count of the number of times the
geolocator was immersed in water as a proxy for esti-
mating the feeding rate across the annual cycle.
We used ANOVA to test if there were differences in
migration parameters (migration timing, speed, dis-
tance, feeding rate, etc.) between males and females.
3
Mar Ecol Prog Ser 638: 1–12, 2020
We used Bartlett’s test of homogeneity of variances
and a Shapiro-Wilk normality test to test homogeneity
of variance and assumption of normality, respectively.
In 3 migration parameters (onset of spring migration,
migration distance in autumn and migration speed in
autumn), homogeneity of variances and assumption
of normality were not met, and in 1 migration parame-
ter (total migration distance), only the assumption of
normality was not met. Thus, we filtered outlying val-
ues from these 4 migration parameters. After exclud-
ing outlying values, total migration distance and mi-
gration speed in autumn met both assumptions. For
onset of spring migration and migration distance in
autumn, we used the nonparametric Wilcoxon rank
sum test. Since no differences were found in any of
the tested migration parameters, we pooled data of
both sexes for all further analyses. Travel speed (km
d−1) is defined here as the total distance travelled di-
vided by the total number of days spent on migration
in each migration season.
2.2.2. Ocean productivity
We used ocean productivity data to compare if
stopover sites of terns were located in areas with
potentially higher food resource abundance as com-
pared to passage areas (migration corridors). We
defined polygons for stationary sites in a radius of 2°
around the median location of the stationary sites
earlier established by the ‘stationary.migration.sum-
mary’ function. Migration corridors were defined as
lines connecting twice-daily positions within the
migration periods with 1° wide buffer around them
and excluding stationary sites.
We downloaded gridded (0.167° grid) weekly
ocean productivity data from the Ocean Productivity
website of the Oregon State University (www.sci-
ence.oregon state. edu/ ocean.productivity/; Behren-
feld & Falkow ski 1997) and extracted productivity
values for all grid cells within the defined stationary
areas and migration corridors corresponding to the
specific weeks. Grid cells that did not contain pro-
ductivity data were omitted from the analyses.
Because ocean productivity data were positively
skewed, we transformed the values using the natural
logarithm.
2.2.3. Wind support
To evaluate the role of wind assistance in facilitat-
ing the migration, we obtained wind data from the
European Centre for Midrange Weather Forecast
(ECMWF; https://www.ecmwf.int/). Wind data were
averaged across the full migration period of the
tracked terns (autumn 2017: 24 Aug−27 Nov, spring
2018: 31 Mar−6 Jun) for each 4×4° grid across the
area between 85° N−85°S and 80° W−35°E. Because
Arctic terns typically migrate at low altitudes near
the water surface (Gudmundsson et al. 1992, Heden-
ström & Åkesson 2016) we used wind measurements
at the surface level. We then used the ‘NCEP.Air-
speed’ function from the R-package ‘RNCEP’ (Kemp
et al. 2012) to calculate wind support for each move-
ment segment (twice-daily positions and movement
direction between them) along the migration routes
of the tracked individuals.
2.2.4. Daylength
We used the light recording data from the geoloca-
tors to estimate the total amount of daylight hours
experienced by individual terns throughout the
annual cycle. Further, we calculated daylight dura-
tion at various latitudes across the year using the R-
package ‘suncalc’ (Thieurmel & Elmarhraoui 2019).
Daylight duration was calculated as the time
between civil dawn and civil dusk when the geomet-
ric centre of the Sun is 6° below the horizon, which
approximately matches light recording sensitivity of
the geolocators’ light sensor.
3. RESULTS
3.1. Migration and non-breeding areas
Terns departed their breeding site in Svalbard in
late August−early September (Box 1) and migrated to
a stopover area in the north Atlantic. Further, 9 indi-
viduals followed the west coast of Africa with later
stopovers in the southeast Atlantic, while 7 individu-
als followed the east coast of South America with
stopover sites in the southwest Atlantic (Fig. 1a).
Interestingly, 2 individuals (BH011 and BH024, both
females) made a loop at the beginning of the south-
bound migration and returned to Svalbard in early
October before continuing their southward move-
ment into tropical latitudes (Fig. S1i,p). Migration
parameters are summarized in Box 1.
At the end of the southbound migration, all tracked
individuals crossed the Antarctic Circle, entering the
24 h daylight zone; thus, location data from late
November until early to mid-February are not avail-
4
Hromádková et al.: Environmental effects on the longest animal migration
able. Between February and April, non-breeding
sites of all but 1 bird were located in the Weddell Sea
(Box 1, Fig. 2). The outlier individual travelled to
the Indian Ocean, residing at ca. 100° E longitude.
Another individual crossed the Drake Passage be -
tween Antarctica and South America in late March,
thus entering the Pacific Ocean before commencing
the northbound migration. Overall, the terns were
highly mobile during the non-breeding period cover-
ing several thousand km and residing at multiple
sites during their 4−5 mo long non-breeding period
in Antarctica (Box 1). Throughout the non-breeding
period, all birds were exposed daily to sub-zero tem-
peratures (Figs. S1 & S2).
Terns started the northbound migration in early
April (Box 1), following an S-shaped migration
pattern through the Atlantic (Fig. 1b). Longer
stopover periods were scarce until the end of the
migration period, when the birds arrived in the
northern Atlantic residing at the same stopover
region as at the beginning of the southbound
migration. During this stopover period, the terns
increased their feeding rate more than 2-fold (as
implied by the number of times the birds were
recorded being in water) as compared to the rest
of the annual cycle (Fig. 3). Travel speed for all
individuals was higher during the northbound
migration compared to the southbound migration
(paired t-test: t15 = −9.56, p < 0.001; Box 1), and
birds arrived back at the breeding colony in late
May−early June, having completed an average
round trip migration distance of 58 500 km (range:
50 200−78 500 km; Box 1).
5
Autumn migration
Start date 3 Sep (24 Aug−11 Sep)
Crossing Equator 4 Oct (23 Sep−22 Oct)
End date 19 Nov (5−27 Nov)
Total migration duration (d) 78 (68−93)
Migration distance (km) 22900 (19500−34800)
Travel speed (km d−1) 294 (258−420)
Non-breeding period
Wintering latitude South of 63° S
Wintering longitude 55° W−94° E
No. days at non-breeding site 137 (127−153)
Movement distance (km) 10812 (5279−19480)
Spring migration
Start date 5 Apr (31 Mar−16 Apr)
Crossing Equator 26 Apr (17 Apr−11 May)
End date 30 May (23 May−6 Jun)
Total migration duration (d) 54 (39−64)
Migration distance (km) 24800 (19600−29700)
Travel speed (km d−1) 435 (334−518)
Total track length (km) 58500 (50200−78500)
Total daylight hours 6985 (6750−7232)
Box 1. Summary data (mean and range) of key migration
and non-breeding period parameters of 16 geolocator-
tracked Arctic terns from a breeding colony in Longyear-
byen, Svalbard
(a) (b)
Fig. 1. Migration routes and stopover areas of 16 geolocator-tracked Arctic terns during (a) southbound and (b) northbound mi-
gration. Breeding site in Longyearbyen, Svalbard, is marked with an orange diamond, stopover sites longer than 3 d are marked
with dots. Background map source: Blue Marble Next Generation, https:// visibleearth. nasa. gov/ collection/ 1484/ blue-marble
Mar Ecol Prog Ser 638: 1–12, 2020
3.2. Migration routes and wind
support
In both seasons, terns on average be -
nefited from tailwind support along their
chosen migration routes (Figs 4 & 5).
During the southbound migration, terns
on average experienced 0.4 ± 0.6 m s−1
(SD) tailwind support, while during the
northbound migration the experienced
tailwind support was substantially
stronger, averaging 2.2 ± 1.2 m s−1
(paired t-test: t15 = −5.34, p < 0.001;
Fig. 5). Testing for the wind support on
reversed migration routes (travelling
along the spring routes in autumn and
autumn routes in spring) revealed that
terns would experience significantly
more headwinds in both migration sea-
sons (southbound migration: −2.1 ± 0.8 m
s−1, paired t-test: t15 = 8.36, p < 0.001;
northbound migration: −0.2 ± 0.7 m s−1,
paired t-test: t15 = 6.81, p < 0.001; Fig. 5).
After excluding the 2 outlying indi-
viduals that made a loop at the begin-
ning of the autumn migration and had
exceptionally high travel speeds, there
was no relationship between wind sup-
port and individual southbound travel
speed (β= −0.72 ± 7.38, F1,12 < 0.01, r2=
0.01, p = 0.924), while there was a posi-
tive relationship between the experi-
6
90°E90°W
180°
0°
8
0
°
S
7
0
°
S
S
6
0
°
t
A
n
a
r
c
t
i
c
i
C
r
c
l
e
5
0
°
S
Fig. 2. Non-breeding areas (February−April) of 16 geolocator-tracked Arctic
terns. Stationary sites where birds remained for at least 7 d are marked with
dots; lines show movements within the non-breeding areas. Location data from
late November (after the southbound migration) until early-mid February are
not available due to 24 h daylight as all birds resided south of the Antarctic
Circle
Jul Aug Sep Oct Nov Dec Jan Feb Mar
Apr May Jun
Jul
Average no. of times in water per day
1000
800
600
400
200
0
Fig. 3. Weekly average (± SD) number of times per day when geolocators were submerged in water during the annual cycle.
Migration periods are marked with grey bars on top (higher colour intensity corresponds to overlapping migration periods of
more individuals) and individual timings of start and end of migration are indicated by open circles
Hromádková et al.: Environmental effects on the longest animal migration 7
enced wind support and individual north-
bound travel speeds (β= 28.89 ± 9.33, F1,14
= 9.6, r2= 0.41, p = 0.008).
3.3. Stopover areas and ocean
productivity
While on migration, terns spent on aver-
age 32 ± 7.9 d (SD) on stopover sites during
the southbound migration and 10 ± 7 d
Wind speed
(m s–1)
10
5
0
0
+5
+10
−5
−10
Wind support
(m s–1)
(a) (b)
Fig. 4. Wind support along the migration routes of 16 geolocator-tracked Arctic terns during (a) southbound and (b) north-
bound migration. Lines show the most likely migration paths coloured according to the wind support in each segment (brown:
headwind, turquoise: tailwind). Directions of arrows indicate gridded wind directions and background heatmap shows wind
speed; both are averaged across the migration period of the tracked Arctic terns. Breeding site in Longyearbyen, Svalbard,
is marked with a black dot
Average wind support (m s–1)
0
−2
−4
2
4
Realised
routes
Reversed
(spring)
routes
Realised
routes
Reversed
(autumn)
routes
(a) (b)
Fig. 5. Average wind support across the entire
migration route for 16 geolocator-tracked Arc-
tic terns during (a) southbound and (b) north-
bound migration. Wind support along the re-
alised migration routes is shown in dark grey
and reversed routes (spring routes in autumn
and vice versa) are shown in light grey. Box
plots show median values with interquartile
ranges (IQR; boxes), whiskers extend to 1.5×
the IQR, outliers are given as open circles
Mar Ecol Prog Ser 638: 1–12, 2020
during the northbound migration.
Stopover sites of the tracked terns were
located in areas where ocean produc-
tivity was higher (mean ± SD; 6.28 ±
0.43 ln[mg C m−2 d−1]) as compared to
migration corridors (6.03 ± 0.39 ln[mg
C m−2 d−1]; paired t-test: t18 = 3.87, p =
0.001; Fig. 6). Weekly locations of
stopover areas and migration corridors
with the underlying ocean productivity
maps can be found in Fig. S3.
3.4. Migration timing and daylight
hours
In both seasons and particularly in
spring, terns on average crossed the
Equator significantly later than the
equinoxes (autumn average: +12 ± 8 d (SD), t-test:
t15 = 6.402, p < 0.001, spring average: +38 ± 6 d; t-test:
t15 = 23.334, p < 0.001; Fig. 7). Despite the low syn-
chronisation between crossing the Equator and
equinoxes, birds still on average experienced 6985 ±
123 h (SD) of daylight during the annual cycle, which
corresponds to 79.7% of all available daylight on
Earth per year (365 days × 24 h = 8760 h).
4. DISCUSSION
In this study, we show that Arctic terns breeding at
78° N in Svalbard migrate to non-breeding sites south
of 63° S in the Weddell Sea covering up to 80 000 km
on their round-trip journeys. This impressive migra-
tion is facilitated by tailwind support along the cho-
sen migration routes in spring, and food-rich stopover
areas for refuelling. Despite the low synchronisation
between Equator crossing and equi noxes, which
would allow for maximizing the experienced daylight
(and thus, the amount of time when birds can forage),
the tracked terns still experienced ca. 80% of all
available daylight hours on the Earth per year, which
is the most by any animal in the world. These findings
suggest that the amount of time when the birds
can feed is not the limiting factor during migration,
but terns rather shift between seasonally specific
8
4.5
5.0
5.5
6.0
6.5
7.0
7.5
Productivity [ln(mg C m–2 day–1)]
OctSep Nov MayApr
Stopover areas
Migration corridors
Southbound migration Northbound migration
Fig. 6. Weekly ocean productivity at stopover areas (black circles) and migration corridors (grey circles) of Arctic terns during
the southbound and northbound migration. Data are mean ± SD
0
80
60
−20
−40
−60
−80
40
Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun
20
Latitude
0
6
12
18
Daylight
(h)
24
Fig. 7. Latitude and experienced daylight duration as a function of time across
the annual cycle indicating the underlying phenological patterns of 16 individ-
ually tracked Artic terns. The dashed parts of the black lines indicate the non-
breeding period when birds were exposed to 24 h polar day making latitude
estimates impracticable. Grey dashed line marks the Equator and shaded ar-
eas indicate periods 2 wk on either side of the equinoxes. Daylight duration
was calculated as the time between civil dawn and civil dusk when the geo-
metric centre of the sun is 6° below the horizon
Hromádková et al.: Environmental effects on the longest animal migration
exploitation of tailwind support (in spring) and ocean
areas of high food abundance (particularly in au -
tumn) to facilitate their remarkable migration.
4.1. Migration, non-breeding areas and population
comparison
In our study, we did not find differences in migra-
tion patterns between males and females, correspon-
ding to earlier studies on trans-Equatorial migratory
seabirds (Shaffer et al. 2006, Guilford et al. 2009,
Magnusdottir et al. 2012, Mosbech et al. 2012). Arctic
terns from Svalbard migrated to their non-breeding
areas in Antarctica via 2 distinct routes following
either the west coast of Africa or the east coast of
South America; both of these routes are known from
earlier studies (González-Solís et al. 2007, Guilford et
al. 2009, Egevang et al. 2010, Fijn et al. 2013, Volkov
et al. 2017, Alerstam et al. 2019). Similarly, the Wed-
dell Sea, where most of our tracked terns over-win-
tered, has previously been established as a prime
non-breeding area for Arctic terns breeding across
various longitudes in the northern hemisphere
(Egevang et al. 2010, Fijn et al. 2013, McKnight et al.
2013, Volkov et al. 2017).
Terns from our study site on average departed
from Svalbard on 3 September, which is from 1 to
2 mo later than previously shown in studies from
more southerly breeding areas (Table S1; Egevang
et al. 2010, Fijn et al. 2013, Loring et al. 2017,
Volkov et al. 2017, Alerstam et al. 2019, Redfern
& Bevan 2020a). The late departure date of our
tracked individuals corresponds with later depar-
ture of other species from northern latitudes
(Butler et al. 1998, Gilg et al. 2013, Davis et al.
2016). As a general pattern, migratory birds breed-
ing at higher latitudes depart from their breeding
areas later than their southern conspecifics, owing
to the later onset of the breeding season at high
latitudes (Conklin et al. 2010, Briedis et al. 2016).
Similarly, our tracked terns arrived at the breeding
sites later than their conspecifics breeding at more
southern latitudes (Egevang et al. 2010, Fijn et al.
2013, Loring et al. 2017, Volkov et al. 2017, Aler-
stam et al. 2019, Redfern & Bevan 2020a). Such
population-level differences in migration timing
can lead to significant variation in migration
strategies, as different populations face different
environmental conditions en route (González-Solís
et al. 2009, Sittler et al. 2011, Hanssen et al. 2016).
In the case of Arctic terns, Alerstam et al. (2019)
suggested that population-specific migration strate-
gies are driven by intraspecific competition and
different costs of migration. Subsequently,
between-population differences in migration timing
lead to population-specific wintering sites. Com-
parative assessment across populations corresponds
with this hypothesis, as terns from Svalbard arrive
in Antarctica relatively late compared to other
populations and over-winter almost exclusively in
the Weddell Sea. Breeding populations from lower
latitudes arrive in Antarctica relatively earlier and
often over-winter further east in the Indian Ocean
(e.g. Alerstam et al. 2019, Redfern & Bevan 2020b).
We found that terns increased their feeding rate
before arrival at the breeding site in spring
(Fig. 3). At least 2 potential explanations for this
may be brought forward: (1) behavioural changes
regarding increased floating on water. However,
using high-frequency data, McKnight et al. (2013)
showed that Arctic terns spend only a small frac-
tion of time floating on water, deeming this an
unlikely explanation. (2) Birds increased their
feeding frequency before arrival at the breeding
site. Earlier studies confirmed Arctic terns as in -
come breeders (species that primarily use local
resources for egg production; Drent & Daan 1980,
Hobson et al. 2000, Mallory et al. 2017); thus, this
increase in feeding rate should not be attributed
to resource deposition for future egg production.
Moreover, we found that both sexes increased
their feeding rate, further ruling out this behaviour
as part of the capital breeding strategy. Another
explanation for the observed changes in feeding
rate may be attributed to changing conditions in
food availability. When food is abundant, birds
may require less time for feeding (e.g. at the
breeding sites), while when food is scarce it may
require more effort and time to feed. However,
ocean productivity data indicate high food avail-
ability at the place and time when the birds
showed increased feeding rate, implying that food
availability may not be the main driver behind the
observed pattern. A more likely explanation may
be preparation against the forthcoming reduction
in feeding time due to courtship behaviour and
incubation. This explanation is also supported by
the reduced number of times when the birds were
in water after their arrival at the breeding sites,
suggesting that birds could at least partially be
using previously stored reserves for body mainte-
nance during egg incubation. Moreover, body
mass of the tagged individuals captured during
incubation showed a decline of 0.75 g d−1over the
capturing period, further supporting this premise.
9
Mar Ecol Prog Ser 638: 1–12, 2020
4.2. Migration patterns and the environment
Our findings suggest that the terns benefit from
using a looped migration strategy where southbound
and northbound migration routes do not overlap. By
adapting the migration routes to the prevailing wind
patterns across the Atlantic Ocean, Arctic terns take
advantage of tailwind support en route. During the
autumn migration, wind support along the 2 main
autumn migration flyways —southeastern and south-
western Atlantic —is essentially different. The over-
all net wind support during the autumn migration
was negligible, and individuals migrating through
the southeastern Atlantic generally experienced
more headwinds. Following this flyway may have a
trade-off between relying on wind support and
fuelling at food-rich stopovers, as ocean productivity
in the southeastern Atlantic during this time of year is
higher compared to the southwestern Atlantic (i.e.
the coasts of Brazil; Fig. S3a−k).
Wind support was particularly pronounced during
the northbound migration when strong tailwinds
likely contributed to the exceptionally fast travel
speed of terns: a 1.5-fold increase compared to travel
speed during the southbound migration (Fig. 4, Box 1;
sensu Kemp et al. 2010). Similar increases in seasonal
travel speed have also been found in other seabirds
(Felicísimo et al. 2008, González-Solís et al. 2009). In
contrast, Hensz (2015) did not find wind to be a signif-
icant predictor for travel speed during either south-
bound or northbound migration of Arctic terns. Wind
exploitation might also be used in combination with a
fly-and-forage strategy (Strandberg & Alerstam 2007)
to further increase travel speed. A fly-and-forage mi-
gration strategy is advantageous when an individual
carries enough energy reserves from the non-breed-
ing grounds and does not need to frequently refuel at
stopovers (Strandberg & Alerstam 2007).
Stopover use, however, seems to play a more
important role in route choice during the southbound
migration. Tracked birds navigated between stop -
over sites of high ocean productivity with migration
corridors passing over areas of lower productivity
(Fig. 6 & Fig. S3). Such patterns in Arctic terns were
first documented by McKnight et al. (2013) as birds
from Alaska migrated along the west coast of the
Americas. Similar patterns were described in a study
where terns were tracked from Greenland and Ice-
land (Hensz 2015). During both migration periods,
one of the main stopover areas of our tracked birds
corresponded with the well-known refuelling area
for migrating seabirds in the North Atlantic (Catry et
al. 2011, Sittler et al. 2011, Gilg et al. 2013), also
known as the North Atlantic drift province
(Longhurst 2010). According to Bourne & Casement
(1996), Arctic terns are present in the area from late
April until late October, with a distinct peak in
August. This time window corresponds with breed-
ing site departure dates of terns from lower (Egevang
et al. 2010, Fijn et al. 2013, Loring et al. 2017, Volkov
et al. 2017) and higher latitudes (our study; Box 1).
Because terns are exclusively diurnal foragers
(McKnight et al. 2013), we predicted that they will
time their migration to cross the Equator close to the
autumnal and vernal equinoxes, thus experiencing
the longest foraging hours (Alerstam 2003). How-
ever, during both migration seasons, terns crossed
the Equator significantly later than the equinoxes,
suggesting that available daylength is not a limiting
factor during migration. The large variation we
observed in Equator crossing dates of the tracked
terns suggests that crossing time is rather flexible.
Furthermore, on southbound migration, Arctic terns
from a breeding site in Sweden crossed the Equator
almost 2 mo earlier than terns in our study (Alerstam
et al. 2019; Table S1). Such population differences
indicate that the timing of crossing the Equator on a
population level is mainly influenced by the timing of
departure from breeding sites rather than by envi-
ronmental conditions.
Collectively, our findings suggest that the slower
southbound migration is primarily guided by com-
muting between food-rich stopovers, whereas the
faster spring migration is adapted to take advantage
of the prevailing wind patterns to facilitate a shorter
migration duration. Disentangling the influence of
environmental drivers behind seasonal migration
strategies of Arctic terns brings us a step closer to
understanding the ecology of the world’s longest ani-
mal migration. Our results provide a means to better
understand the delicate relationship between sea-
sonal migration strategies of birds and variation in
environmental conditions, which may be disrupted
by the ongoing global climate change.
Acknowledgements. We thank all of our field assistants for
their help in locating and capturing birds, Simeon Lisovski
for help with processing wind data, and Radka Piálková for
guidance in the lab. We also thank the Czech Arctic Scien-
tific Infrastructure of the University of South Bohemia in
>eské Budeˇ jovice − the Josef Svoboda Station in Svalbard
(CzechPolar2 project LM2015078 supported by the Ministry
of Education, Youth and Sports of the Czech Republic) for
housing us during the field season. Financial support was
provided by the Palacky University Endowment Fund (to
M.B.), by the Ministry of Education, Youth and Sports of the
Czech Republic (LM2015078 to T.H. and V.P.), Grant
10
Hromádková et al.: Environmental effects on the longest animal migration
Agency of the University of South Bohemia (GAJU n. 04-
151/2016/P and n. 048/2019/P to T.H.), The Explorers Club
Exploration Fund Grant (to T.H.) and the Latvian Council
of Science (LZP/2019/45 to M.B.). All experiments were
conducted according to Norwegian law. The study was
approved by the Norwegian Food Safety Authority (ref.
number 19/69972) and by the Governor of Svalbard (ref.
number 17/00693-8) and is registered in the Research in
Svalbard database (RiS ID 10805).
LITERATURE CITED
Alerstam T (2003) Bird migration speed. In: Berthold P,
Gwinner E, Sonnenschein E (eds) Avian migration.
Springer, Berlin, p 253−267
Alerstam T (2011) Optimal bird migration revisited. J
Ornithol 152: 5−23
Alerstam T, Lindström Å (1990) Optimal bird migration: the
relative importance of time, energy, and safety. In: Gwin-
ner E (ed) Bird migration. Springer, Berlin, p 331−351
Alerstam T, Bäckman J, Grönroos J, Olofsson P, Strandberg
R (2019) Hypotheses and tracking results about the
longest migration: the case of the arctic tern. Ecol Evol 9:
9511−9531
Bauchinger U, Klaassen M (2005) Longer days in spring than
in autumn accelerate migration speed of passerine birds.
J Avian Biol 36: 3−5
Behrenfeld MJ, Falkowski PG (1997) A consumer’s guide to
phytoplankton primary productivity models. Limnol
Oceanogr 42: 1479−1491
BirdLife International (2018) Sterna paradisaea. The IUCN
Red List of Threatened Species 2018: e. T22694629A1320
65195. https: //dx.doi.org/10.2305/IUCN.UK.2018-2. RLTS.
T22694629A132065195.en (accessed 25 January 2020)
Bourne WRP, Casement MB (1996) The migrations of the
Arctic tern. Bull Br Ornithol Club 116: 117−122
Briedis M, Hahn S, Gustafsson L, Henshaw I, Träff J, Král M,
Adamík P (2016) Breeding latitude leads to different
temporal but not spatial organization of the annual cycle
in a long-distance migrant. J Avian Biol 47: 743−748
Brlík V, Kole<ek J, Burgess M, Hahn S and others (2020)
Weak effects of geolocators on small birds: a meta-analy-
sis controlled for phylogeny and publication bias. J Anim
Ecol 89: 207−220
Butler PJ, Woakes AJ, Bishop CM (1998) Behaviour and
physiology of Svalbard barnacle geese Branta leucop-
sis during their autumn migration. J Avian Biol 29:
536−545
Catry P, Dias MP, Phillips RA, Granadeiro JP (2011) Differ-
ent means to the same end: long-distance migrant sea-
birds from two colonies differ in behaviour, despite com-
mon wintering grounds. PLOS ONE 6: e26079
Conklin JR, Battley PF, Potter MA, Fox JW (2010) Breeding
latitude drives individual schedules in a trans-hemi-
spheric migrant bird. Nat Commun 1: 67
Davis SE, Maftei M, Mallory ML (2016) Migratory connec-
tivity at high latitudes: Sabine’s gulls (Xema sabini) from
a colony in the Canadian high arctic migrate to different
oceans. PLOS ONE 11: e0166043
Devlin CM, Diamond AW, Saunders GW (2004) Sexing Arc-
tic terns in the field and laboratory. Waterbirds 27:
314−320
Drent R, Daan S (1980) The prudent parent: energetic
adjustments in avian breeding. Ardea 68: 225−252
Egevang C, Stenhouse IJ, Phillips RA, Petersen A, Fox JW,
Silk JRD (2010) Tracking of Arctic terns Sterna paradis-
aea reveals longest animal migration. Proc Natl Acad Sci
USA 107: 2078−2081
Felicísimo ÁM, Muñoz J, González-Solis J (2008) Ocean sur-
face winds drive dynamics of transoceanic aerial move-
ments. PLOS ONE 3: e2928
Fijn RC, Hiemstra D, Phillips RA, van der Winden J (2013)
Arctic terns Sterna paradisaea from The Netherlands
migrate record distances across three oceans to Wilkes
Land, East Antarctica. Ardea 101: 3−12
Fletcher KL, Hamer KC (2003) Sexing terns using biomet-
rics: the advantage of within-pair comparisons. Bird
Study 50: 78−83
Fridolfsson AK, Ellegren H (1999) A simple and universal
method for molecular sexing of non-ratite birds. J Avian
Biol 30: 116−121
Gilg O, Moe B, Hanssen SA, Schmidt NM and others (2013)
Trans-equatorial migration routes, staging sites and win-
tering areas of a high-arctic avian predator: the long-
tailed skua (Stercorarius longicaudus). PLOS ONE 8:
e64614
González-Solís J, Croxall JP, Oro D, Ruiz X (2007) Trans-
equatorial migration and mixing in the wintering areas of
a pelagic seabird. Front Ecol Environ 5: 297−301
González-Solís J, Felicísimo A, Fox JW, Afanasyev V, Kol-
beinsson Y, Muñoz J (2009) Influence of sea surface
winds on shearwater migration detours. Mar Ecol Prog
Ser 391: 221−230
Green M (2004) Flying with the wind – spring migration of
arctic-breeding waders and geese over south Sweden.
Ardea 92: 145−160
Griffiths R, Double MC, Orr K, Dawson RJG (1998) A DNA
test to sex most birds. Mol Ecol 7: 1071−1075
Gudmundsson GA, Alerstam T, Larsson B (1992) Radar ob -
servations of northbound migration of the Arctic tern,
Sterna paradisaea, at the Antarctic Peninsula. Antarct
Sci 4: 163−170
Guilford T, Meade J, Willis J, Phillips RA and others (2009)
Migration and stopover in a small pelagic seabird, the
Manx shearwater Puffinus puffinus: insights from
machine learning. Proc R Soc B 276: 1215−1223
Hanssen SA, Gabrielsen GW, Bustnes JO, Bråthen VS and
others (2016) Migration strategies of common eiders from
Svalbard: implications for bilateral conservation man-
agement. Polar Biol 39: 2179−2188
Hedenström A, Åkesson S (2016) Ecology of tern flight in
relation to wind, topography and aerodynamic theory.
Philos Trans R Soc B 371: 20150396
Hedenström A, Alerstam T (1997) Optimum fuel loads in
migratory birds: distinguishing between time and energy
minimization. J Theor Biol 189: 227−234
Hedenström A, Alerstam T (1998) How fast can birds
migrate? J Avian Biol 29: 424−432
Hensz CM (2015) Environmental factors in migratory route
decisions: a case study on Greenlandic Arctic terns
(Sterna paradisaea). Anim Migr 2: 76−85
Hobson KA, Sirois J, Gloutney ML (2000) Tracing nutrient
allocation to reproduction with stable isotopes: a prelim-
inary investigation using colonial waterbirds of Great
Slave Lake. Auk 117: 760−774
Kemp MU, Shamoun Baranes J, Van Gasteren H, Bouten W,
Van Loon EE (2010) Can wind help explain seasonal dif-
ferences in avian migration speed? J Avian Biol 41:
672−677
11
Mar Ecol Prog Ser 638: 1–12, 2020
Kemp MU, Shamoun-Baranes J, van Loon EE, McLaren JD,
Dokter AM, Bouten W (2012) Quantifying flow-assis-
tance and implications for movement research. J Theor
Biol 308: 56−67
Kranstauber B, Weinzierl R, Wikelski M, Safi K (2015)
Global aerial flyways allow efficient travelling. Ecol Lett
18: 1338−1345
Kvist A, Lindström Å (2000) Maximum daily energy intake:
It takes time to lift the metabolic ceiling. Physiol Biochem
Zool 73: 30−36
Ležalová-Piálková R (2011) Molecular evidence for extra-
pair paternity and intraspecific brood parasitism in the
black-headed gull. J Ornithol 152: 291−295
Lindström Å (2003) Fuel deposition rates in migrating birds:
causes, constraints and consequences. In: Berthold P,
Gwinner E, Sonnenschein E (eds) Avian migration.
Springer, Berlin, p 307−320
Lisovski S, Bauer S, Briedis M, Davidson SC and others
(2020) Light-level geolocator analyses: a user’s guide. J
Anim Ecol 89: 221−236
Longhurst AR (2010) Ecological geography of the sea, 2nd
edn. Academic Press, San Diego, CA
Loring P, Ronconi R, Welch L, Taylor P, Mallory M (2017)
Postbreeding dispersal and staging of common and Arc-
tic terns throughout the western North Atlantic. Avian
Conserv Ecol 12: 20
Magnusdottir E, Leat EHK, Bourgeon S, Strøm H and others
(2012) Wintering areas of great skuas Stercorarius skua
breeding in Scotland, Iceland and Norway. Bird Study
59: 1−9
Mallory ML, Boadway KA, Davis SE, Maftei M, Diamond
AW (2017) Breeding biology of Arctic terns (Sterna par-
adisaea) in the Canadian High Arctic. Polar Biol 40:
1515−1525
McKnight A, Allyn AJ, Duffy DC, Irons DB (2013) ‘Stepping
stone’ pattern in Pacific Arctic tern migration reveals the
importance of upwelling areas. Mar Ecol Prog Ser 491:
253−264
Mosbech A, Johansen KL, Bech NI, Lyngs P and others
(2012) Inter-breeding movements of little auks Alle alle
reveal a key post-breeding staging area in the Green-
land Sea. Polar Biol 35: 305−311
R Core Team (2018) R: a language and environment for sta-
tistical computing. R Foundation for Statistical Comput-
ing, Vienna
Rakhimberdiev E, Saveliev A, Piersma T, Karagicheva J
(2017) FLightR: an R package for reconstructing animal
paths from solar geolocation loggers. Methods Ecol Evol
8: 1482−1487
Redfern CPF, Bevan RM (2020a) Overland movement and
migration phenology in relation to breeding of Arctic
terns Sterna paradisaea. Ibis 162: 373–380
Redfern CPF, Bevan RM (2020b) Use of sea ice by arctic
terns Sterna paradisaea in Antarctica and impacts of cli-
mate change. J Avian Biol 51: e02318
Shaffer SA, Tremblay Y, Weimerskirch H, Scott D and others
(2006) Migratory shearwaters integrate oceanic
resources across the Pacific Ocean in an endless summer.
Proc Natl Acad Sci USA 103: 12799−12802
Sittler B, Aebischer A, Gilg O (2011) Post-breeding migra-
tion of four long-tailed skuas (Stercorarius longicaudus)
from North and East Greenland to West Africa. J
Ornithol 152: 375−381
Storr GM (1958) Migration routes of the Arctic tern. Emu 58:
59−62
Strandberg R, Alerstam T (2007) The strategy of fly-and-for-
age migration, illustrated for the osprey (Pandion haliae-
tus). Behav Ecol Sociobiol 61: 1865−1875
Thieurmel B, Elmarhraoui A (2019) Suncalc: compute sun
position, sunlight phases, moon position and lunar phase.
R Package version 050, https:// CRAN.R-project.org/
package = suncalc
Volkov AE, Loonen MJJE, Volkova EV, Denisov DA (2017)
New data for Arctic terns (Sterna paradisaea) migration
from White Sea (Onega peninsula). Ornithologia 41:
58−68
12
Editorial responsibility: Stephen Wing,
Dunedin, New Zealand
Submitted: October 31, 2019; Accepted: February 25, 2020
Proofs received from author(s): March 13, 2020
