ArticlePDF Available

Abstract and Figures

The migratory movements of seabirds (especially smaller species) remain poorly understood, despite their role as harvesters of marine ecosystems on a global scale and their potential as indicators of ocean health. Here we report a successful attempt, using miniature archival light loggers (geolocators), to elucidate the migratory behaviour of the Manx shearwater Puffinus puffinus, a small (400 g) Northern Hemisphere breeding procellariform that undertakes a trans-equatorial, trans-Atlantic migration. We provide details of over-wintering areas, of previously unobserved marine stopover behaviour, and the long-distance movements of females during their pre-laying exodus. Using salt-water immersion data from a subset of loggers, we introduce a method of behaviour classification based on Bayesian machine learning techniques. We used both supervised and unsupervised machine learning to classify each bird's daily activity based on simple properties of the immersion data. We show that robust activity states emerge, characteristic of summer feeding, winter feeding and active migration. These can be used to classify probable behaviour throughout the annual cycle, highlighting the likely functional significance of stopovers as refuelling stages.
Content may be subject to copyright.
Migration and stopover in a small pelagic seabird,
the Manx shearwater Puffinus puffinus: insights
from machine learning
T. Guilford
1,
*, J. Meade
2
, J. Willis
3
, R. A. Phillips
4
, D. Boyle
5
, S. Roberts
3
,
M. Collett
1
, R. Freeman
6
and C. M. Perrins
7
1
Animal Behaviour Research Group, Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK
2
Department of Animal and Plant Sciences, University of Sheffield, Western Bank, Sheffield S10 2TN, UK
3
Department of Engineering Science, University of Oxford, Parks Road, Oxford OX1 3PJ, UK
4
British Antarctic Survey, Natural Environment Research Council, High Cross, Madingley Road, Cambridge CB3 0ET, UK
5
Skomer Island National Nature Reserve, Marloes, Nr. Haverfordwest, Pembrokeshire SA62 3BL, UK
6
Computational Ecology and Environmental Science, Microsoft Research, JJ Thompson Avenue, Cambridge CB3 0FB, UK
7
Edward Grey Institute of Field Ornithology, Department of Zoology, University of Oxford, South Parks Road,
Oxford OX1 3PS, UK
The migratory movements of seabirds (especially smaller species) remain poorly understood, despite their
role as harvesters of marine ecosystems on a global scale and their potential as indicators of ocean health.
Here we report a successful attempt, using miniature archival light loggers (geolocators), to elucidate the
migratory behaviour of the Manx shearwater Puffinus puffinus, a small (400 g) Northern Hemisphere
breeding procellariform that undertakes a trans-equatorial, trans-Atlantic migration. We provide details of
over-wintering areas, of previously unobserved marine stopover behaviour, and the long-distance
movements of females during their pre-laying exodus. Using salt-water immersion data from a subset of
loggers, we introduce a method of behaviour classification based on Bayesian machine learning techniques.
We used both supervised and unsupervised machine learning to classify each bird’s daily activity based on
simple properties of the immersion data. We show that robust activity states emerge, characteristic of
summer feeding, winter feeding and active migration. These can be used to classify probable behaviour
throughout the annual cycle, highlighting the likely functional significance of stopovers as refuelling stages.
Keywords: geolocator tracking technology; tracking pelagic seabirds; Bayesian machine learning;
migration; Manx shearwater; spatial ecology
1. INTRODUCTION
The long-distance migratory movements and winter
habitats of pelagic seabirds remain poorly understood,
despite these animals’ role as harvesters of marine
ecosystems on a global scale and their potential as
indicators of ocean health (Shaffer et al. 2006). Traditional
studies based on ringing recoveries or ocean sightings have
proved valuable for identifying very general movement
patterns, but fail to characterize any behavioural detail or
discriminate important localities at sea, which may be
critical both for understanding the dynamics of oceanic
migration processes, and for identifying Important Bird
Areas and candidate Marine Protection Area networks for
conservation. New telemetry systems continue to revolu-
tionize the study of elusive pelagic lifestyles, so that the
foraging ranges and migratory behaviour of larger,
predominantly southern species such as albatrosses and
larger petrels are becoming much better understood
(Weimerskirch et al. 2002; Phillips et al. 2006,2008).
Recently (Shaffer et al. 2006;Gonzales-Solis et al. 2007;
Felicisimo et al. 2008), miniature geolocation technology
has been used to track sub-1000 g seabirds on trans-
equatorial oceanic migrations. In this study, we used
geolocation technology, combined with a novel appli-
cation of analytical techniques adapted from machine
learning, to elucidate the details of large-scale pelagic
movements in a 400 g Northern Hemisphere breeding
procellariform, the Manx shearwater Puffinus puffinus,
both during breeding and during its trans-equatorial,
trans-Atlantic annual migration.
The majority of the world population of Manx shear-
waters breed in Britain and Ireland ( Hamer 2003),
combining spatially restricted colonial breeding with a
highly pelagic lifestyle. The species is therefore of
particular interest to the study of migration ecology, and
potentially to understanding the impacts of changing
ocean health on a global integrator of marine resources.
Much of the breeding biology is known from the studies on
two Welsh islands, Skokholm and Skomer (e.g. Brooke
1990), which hold in excess of 150 000 breeding pairs
(Smith et al. 2001), perhaps 35–40 per cent of the total
breeding population (Hamer 2003). Recently, miniature
GPS devices have enabled us to track breeding birds on
their feeding trips from the colony (Guilford et al. 2008)
but, as with most pelagic seabirds, very little detail is
Proc. R. Soc. B (2009) 276, 1215–1223
doi:10.1098/rspb.2008.1577
Published online 13 January 2009
Electronic supplementary material is available at http://dx.doi.org/10.
1098/rspb.2008.1577 or via http://journals.royalsociety.org.
*Author for correspondence (tim.guilford@zoo.ox.ac.uk).
Received 31 October 2008
Accepted 11 December 2008 1215 This journal is q2009 The Royal Society
known about their behaviour or habitat requirements
during the non-breeding season (September to March).
Ringing recoveries indicate that Manx shearwaters spend
the northern winter off the coast of South America, with
most recoveries coming from Brazil (Perrins & Brooke
1976;Thompson 2002;Hamer 2003), but such data do
not provide accurate wintering destinations and provide
almost no behavioural detail. Furthermore, females under-
take a pre-laying exodus (Brooke 1990) during which they
build their large egg (15% of body weight), a critically
important behavioural characteristic of procellariforms that
remains poorly understood in this or any species.
In addition to geolocation data, a subset (7 out of 12) of
our devices collected salt-water immersion data, logging
the proportion of every 10 min period that the bird was on
or under the water as opposed to in the air. We used a
Bayesian analysis adapted from machine learning tech-
niques to identify distinct behavioural categories inherent
in the patterns within the immersion records, and used
this to shed light on the birds’ behaviour at different stages
of the migratory cycle: during summer feeding, winter
feeding, migration and egg formation. In particular, we
identify what appear to be migratory stopover periods that
we hypothesize may function in the same way as stopovers
in terrestrial migrants for refuelling. We believe that this
and similar machine learning techniques may have
considerable usage in helping to extract more information
from extant and future animal tracking datasets.
2. MATERIAL AND METHODS
(a)Geolocators
We fitted twelve 2.5 g archival light logging geolocators to
elliptically shaped plastic rings to the legs of both members of
six established breeding pairs on Skomer Island, Wales
(coordinates 518440N, 58190W). The five Mk6 and seven
Mk9 devices, designed and manufactured by the British
Antarctic Survey, Cambridge, were ground-truthed before
and after deployment, deployed in late August 2006, and
retrieved after egg laying in the following breeding season. All
12 birds returned to breed, and data were retrieved
successfully from all 12 devices.
(b)Analysis of location data
Data were analysed initially in MULTITRACE software ( Jensen
Software Systems), with the correction factor for day/night
movement set to 0.5, an elevation angle of K5.58and a simple
correction for fast east/west movement (see MT Geolocation
manual available at: http://www.jensen-software.com/
downloads.html July 2008 for more details), a choice aided
by reference to the known ground-truthed position at the
colony. Day length is used to provide an estimate of latitude,
while the timing of recorded midday or midnight is used to
provide an estimate of longitude. The quality of our light
curves was very good, so it is likely that error is similar to that
found in previous studies even at higher latitudes (see Phillips
et al. 2004 and Shaffer et al. 2005 for discussion of geolocation
accuracy). Because day length approaches uniformity across
the globe at the equinoxes, erroneous location estimates are
often seen around these periods and these were excluded.
Points were also regarded as erroneous if they involved more
than 1600 km position change in 1 day, lay in a line along the
equator or were affected by obvious interference in the light
curve around sunrise and sunset. Eighty per cent occupancy
kernels were used to identify individual over-wintering areas,
and a 500 km boundary from the colony was used to define
the breeding area, for the purposes of calculating migration,
breeding or over-wintering statistics. Since longitude is more
accurately estimated than latitude (e.g. Wilson et al. 1992;
Hill 1994), especially around the equinoxes, we used change
in longitude to identify stopover periods during migration.
A stopover was defined as a period in which there was less
than a 0.88change in longitude in half a day smoothed over
3 days and contiguous for at least 2.5 days.
(c)A Bayesian approach to inferring behaviour
classes from immersion data
All seven Mk9 devices returned with complete logs of salt-
water immersion. Each device registered whether or not it
was submerged in salt-water every 3 s, and recorded the
number of such immersions in each 10 min period through-
out every day as a value from 0 (in air over entire period) to
200 (immersed over entire period). Figure 1 illustrates how
immersion data are distributed over a typical 24 hour period.
These data can be used simply to determine whether a bird is
flying, or is on or under the water. However, since the pattern
of immersion versus flying events across a day or night is likely
to reflect a bird’s behaviour on a larger scale (e.g. foraging
versus long- or short-distance movements), we attempted to
determine whether consistent patterns existed in the data.
Our approach was to develop a method of unsupervised
probabilistic machine classification, which we then cross-
verified with a supervised classification based on prior
knowledge of likely behavioural states (further explanation
is given in the electronic supplementary material). First, we
determined a set of candidate dimensions for each 24 hour
period, such as number of wet events (defined as crossings
over a threshold value of 100; see figure 1), number of dry
events or the mean time of all wet events during the day or
night (time over or under threshold). We chose three
dimensions that were well-distributed over their range to
allow the classifier to subdivide the space most effectively into
separate classes. We made no assumption about the order of
the days, the particular birds or the meaning of the behaviour.
As the best dimensions for the unsupervised classifier, we
chose: (i) mean length of dry events under the threshold in
daytime, (ii) total wet time as a proportion of daytime, and
(iii) total wet time as a proportion of night time. We used a
Gaussian mixture model for unsupervised classification with
a number of components determined from the data and
inferred using the variational Bayes learning approach (Attias
2000;Roberts & Penny 2002;Roberts et al. 2004;Bishop
2006). The unsupervised classifier automatically clusters data
according to its self-similarity into a number of putative
groupings and, for each 24 hour section, gives a probability of
membership to one of a set of putative classes. This classifier
(which starts with 20 potential classes, and pares down to the
most parsimonious subset) produced three major classes and
several minor classes (98% of days were classified into three
main classes), and the classifications were similar across
all birds.
Next, we used a supervised classification ( Fisher 1936;
Martinez & Martinez 2008) to provide some verification of
our unsupervised classification (figure 2). To train the
classifier, we divided broad sections of the data into
migration, summer feeding and over-wintering for an
arbitrarily chosen bird, using a combination of dates,
known activity at the colony and location as revealed by the
1216 T. Guilford et al. Pelagic seabird movements
Proc. R. Soc. B (2009)
light data. The same selection of candidate dimensions within
the immersion data used for the unsupervised classifier was
again examined using the supervised classifier. Three
dimensions were chosen that best differentiated between
classes after direct comparison of probability density
functions of class membership for the training data.
200
180
160
140
120
100
80
60
40
20
0 204060
ten minute intervals in 24 h
salt water immersion (no. of 3 s periods in 10 min)
80 100 120 140
Figure 1. Example of salt-water immersion data for a 24 hour period for bird FR53237 on 28 July 2006. Dark (night time)
periods are shown shaded grey and a threshold at 100 units is indicated by a grey line. The plot shows how the bird changes
frequently between sitting/diving and flying/on land, with a period of flight in the middle of the day (immersion value 0), and
approximately 100 min period from nightfall sitting on the water (immersion value 200), presumably rafting, before coming
ashore (immersion value 0).
–100
latitude
–50
0
50
100(a)
(b)
–100 Au
g
2006
longitude for middle points
Se
p
2006 Nov 2006 Jan 2007 Feb 2007 A
p
r 2007 Jun 2007
–80
–60
–40
–20
0
20
Figure 2. (a) Unsupervised classification from immersion data for FP52700 overlaid on latitude. The likely relative error in
latitude is shown in grey, and is calculated by making perturbations around the measured value. With a constant precision in
measurement, error in latitude near the equinox is higher because the day length is similar for all latitudes (i.e. perturbations
have to be greater to produce a similar effect than at other times of year). (b) Supervised classification from immersion data
(upper trace) and unsupervised classification from immersion data (middle trace) overlaid on longitude are shown for the same
bird as in (a). The lower trace indicates periods of migratory movement defined entirely from longitudinal displacement (the
longitude positions were smoothed with a boxcar filter of length 3 days and a position classified as migration if it differed from
the previous one by more than 0.8 degrees). The upper and lower traces have been offset by G208for clarity. The general
agreement between the classifiers is shown, as are several stopovers on the outward- and inward-bound migrations that are
independently classified as feeding from the immersion data. Blue dots, winter feeding; green dots, migration; red dots, summer
feeding; black dots, feeding (either); grey line, position; grey area, position error.
Pelagic seabird movements T. Guilford et al. 1217
Proc. R. Soc. B (2009)
We used odd and even alternative data points as training
groups to provide classification for the training data within
the initially classified periods chosen by the expert, as well as
at times which had been initially unclassified. This method
allows for cross-validation to calculate the probability of
correct classification (which was 0.79; Martinez & Martinez
2008). Supervised classification produces a probability of
membership for each class. Each data point (24 hour period)
is assigned to the class it has the highest probability of
belonging to. A lower threshold was also set where the
probability of membership of any class was so low that the
point remained unclassified. All days from all birds were
classified using these same training data from the arbitrarily
chosen bird. The classification of activity was remarkably
consistent across all birds and compared well in all cases with
the output of the unsupervised classifier. We made a
comparison by matching the class assignments from the two
classifiers and, for each bird, calculating the percentage of points
that were consistently classified. There was a mean consistent
classification of 67 per cent (s.d. 5%) for the seven birds.
3. RESULTS
(a)Broad movement patterns
Figure 3 summarizes the broad pattern of migration for all
12 birds. It is important to recognize that, while clearly
erroneous points based on putative movement distances
have been removed, points on land have not because of the
biasing effect that this procedure would have on overall
distribution centres. This is because there is no a priori
reason to expect land-based points to be less accurate than
those over the ocean. Nevertheless, known natural history
of this species makes it extremely unlikely that such points
represent genuine excursions onto land. Inferential gaps
still exist because of poor resolution around the autumn
equinox in particular, but the data indicate a southwards
migration down the west coast of Africa and across to the
Brazilian coast via approximately the shortest route, then
south or southwest to over-wintering quarters on the
Patagonian Shelf off Argentina centred on latitude 408
south. Fifty per cent occupancy contours at the over-
wintering area suggest that individual over-wintering cores
50°N
25°N
25°S
50°S
80°W60°W40°W70°W50°W60°W
0
48°S
44°S
40°S
36°S
500 km
Rio de la Plata 32°S
Figure 3. Positions of 12 shearwaters tracked with geolocators. Each bird is represented by a different colour. Coloured lines
serve to connect the positions in series providing approximate trajectories. However, where erroneous locations have been
excluded, lines may sometimes connect neighbouring positions that are many days apart and hence are not indicative of actual
routes travelled (e.g. over land). For clarity around the breeding colony and the main over-wintering area (inside the dashed
boxes) where there is a high density of points, plots are of mean positions over two-week periods. The inset shows the 50%
occupancy contours within the southern dashed box around the over-wintering area for all daily positions within that box, using
the same colour scheme as the tracks. Bathymetry contours at 1000 m intervals indicate the edge of the Patagonian Shelf.
1218 T. Guilford et al. Pelagic seabird movements
Proc. R. Soc. B (2009)
are quite similar for all 12 birds, although with the possi-
bility that there may be two distinct destinations. All birds
overwintered close to the Argentinean coast. The return
northwards tended to follow a westwardly curved route
through the eastern Caribbean, even as far as the eastern
seaboard of the US and back through the north Atlantic.
A single (female) bird flew a more direct return migration.
Beyond the broad pattern, however, several important
details emerge (see table 1 for values and statistics). Males
and females departed on autumn migration (defined as
date last recorded within 500 km of the colony) at about
the same time, and migration to the 80 per cent over-
wintering kernel did not cover significantly different
periods (although a larger sample might reveal
differences). The duration of the return (northerly)
migration, however, was significantly longer in males
than females. This was the case even though males tended
to arrive on the colony earlier than females (although not
significantly so in our sample). The fastest sustained
migratory travel was recorded for a male during his southerly
migration when he covered, within the uncertain accuracy
limits of the devices, a straight-line distance of approxi-
mately 7750 km in 6.5 days, a speed of 1192 km a day.
In fact, using the immersion data, we were able to calculate
that this bird achieved this journey in 139 hours airtime,
hence an average surface flight speed of 55 km h
K1
.
(b)Stopovers
Birds did not migrate continuously. Instead, they often
split their journeys with one or more periods in which
there was little onward progress. Such periods are poorly
resolved spatially during the early part of the outward
migration close to the autumn equinox; but during the
latter part of outward migration and during return
migration, birds appeared to spend from a few days to as
long as two weeks in activity apparently unrelated to
migratory travel (figure 4). If we define these periods as
stopovers using minimal change in longitude, we see that
all birds made stopovers, that they are made by both males
and females, and that they are made with approximately
Table 1. Summary of migration statistics.
males (nZ6) females (nZ6) Wilcoxon rank-sum test
southerly migration departure date,
mean (and range)
20 Sep (07 Sep–23 Sep) 21 Sep (13 Sep–23 Sep) pZ1.000
southerly migration duration, mean
(and range)
25.8 days (14–44 days) 35.0 days (26–42 days) pZ0.092
time at over-wintering core (80%),
mean (and range)
139.7 days (125–150 days) 149.3 days (138–156 days) pZ0.058
northerly migration departure date,
mean (and range)
3 Mar (08 Feb–30 Mar) 23 Mar (01 Mar–07 Apr) pZ0.065
northerly migration duration, mean
(and range)
40.2 days (30–58 days) 27.6 days (22–35 days) pZ0.019
arrival near colony, mean
(and range)
13 Apr (06 Apr–29 Apr) 20 Apr (23 Mar–06 May) pZ0.229
stopover days southwards, mean
(and range)
8.1 days (0–30.5 days) 13.9 days (0–23 days) pZ0.309
stopover days northwards, mean
(and range)
13.3 days (0–23 days) 7.4 days (0–18.5 days) pZ0.143
number of stopovers per bird, mean
(and range)
3.2 (1–6) 3.3 (2–5) pZ0.871
duration of individual stopovers,
mean (and range)
6.7 days (2.5–13.5 days) 6.4 days (2.5–15.5 days) pZ0.516
total number of stopover days, mean
(and range)
21.3 days (4.5–41.5 days) 21.3 days (14.5–32 days) pZ0.872
60°N
40°N
20°N
20°S
40°S
80°W60°W40°W20°W0°
0°
Figure 4. Stopovers, sized in proportion to the length of the
stop, are shown at mean location of positions. Blue, male; red,
female (small circles, duration 5 days; big circles, duration
10 days). Outward-bound stops from colony towards winter
feeding ground are indicated with a cross. Locations
apparently over land almost certainly do not indicate that
birds were stopping inland, but serve to emphasize that
position estimates may be subject to considerable error.
Pelagic seabird movements T. Guilford et al. 1219
Proc. R. Soc. B (2009)
equal frequency and duration on both outward and return
migrations (table 1). We hypothesize that these are marine
stopovers functionally equivalent to the stopovers charac-
teristic of many terrestrially migrating birds replenishing
their reserves (see Newton 2008 for a recent review).
(c)Pre-laying exodus
Using light and/or immersion data, we were able to
identify the egg-formation period by examining the male
and female simultaneous presence on land at the colony at
the start of the breeding season followed by a protracted
period of female absence. For some pairs, approximate or
even exact egg-laying dates from burrow inspections were
used as verification (pre-laying exodus: mean start dateZ3
May; mean end dateZ23 May; mean durationZ20 days).
Figure 5 shows the approximate tracks and positions of
the six breeding pairs with sexes plotted separately.
(d)Behaviour classification using immersion data
Three dominant classes emerge robustly from the
unsupervised classification of the salt-water immersion
data. Assigning these three classes to data on plots of
latitude or longitude indicates that they generally corre-
spond to definable periods of the shearwater’s year
(figure 2). One class is most often associated with major
changes in position (green dots, figure 2), and the other
two classes are associated predominantly with activity
during the breeding season (red dots) and during the over-
wintering period (blue dots).
Plotting these states on a spatial map provides a clearer
picture of the shearwater’s activity during the migratory
cycle. Figure 6 shows the tracks of four birds, two male
and two female, with successive points joined by lines
(noting that there may be periods of missing data between
two successive points, particularly around the equinoxes).
Colours refer to membership of the most probable class
identified by the unsupervised classification. Since this
classification shows a strong (67%) overlap with the
supervised classification, we may use the terms summer
feeding (red dots, figure 6), winter feeding (blue dots) and
migration (green dots) to label the classes.
For the pre-laying exodus period immersion data,
classification was possible for four out of the six female
birds. For all the classifiable days (147) using both
classifiers, 56 per cent of days were classified as winter
feeding and 32 per cent as summer feeding, with the
remainder classified as migration (although here the
dataset is small and the congruence between the two
classifiers is less good).
4. DISCUSSION
In contrast to two previously studied trans-equatorial
migrant shearwaters, Cory’s shearwater Calonectris diomedea
and sooty shearwater Puffinus griseus, which exploit several
different over-wintering areas even when birds are from the
same colony (Shaffer et al. 2006;Gonzales-Solis et al. 2007),
breeding Manx shearwaters appear to depend on a single,
remarkably restricted area (with the caveat that this and
other inferences are from data gathered on 12 birds). This
area is close to the Argentinean coast south of the Rı
´odela
Plata, further south than expected from ringing recoveries
(Thompson 2002), but on an area of the Patagonian Shelf
where mixing ocean currents are known to produce rich
fisheries heavily exploited both by other pelagic animals and
by humans (e.g. Phillips et al. 2005).
The broad pattern of migration to and from this over-
wintering area is consistent with the earlier models based
on indirect data, primarily ringing recoveries (Brooke
1990). It suggests a considerable degree of control by
oceanic winds, as has been argued for this and other
species by Felicisimo et al. (2008). The distinctly different
outward and return routes north of the equator suggest
reliance on the trade winds of the north Atlantic gyre. As
noted above, the fastest sustained migratory travel was
recorded as approximately 55 km h
K1
for a male in flight
for a continuous 139 hours. This is the maximum range
speed estimated for this species by Pennycuick (1969), but
it is considerably faster than the 40 km h
K1
mean airspeed
estimate measured during foraging excursions of Manx
shearwaters using GPS (Guilford et al. 2008), which
implies that such performance is likely to have been
achieved with considerable use of favourable wind
16°W
56°N
54°N
52°N
50°N
48°N
46°N
12°W8°W4°W0°16°W
56°N(a)(b)
54°N
52°N
50°N
48°N
46°N
12°W8°W4°W0°
300 km
Figure 5. Positions averaged over 2-day periods for six shearwater pairs, (a) male and (b) female, during the period immediately
after the return from winter migration and before the egg is laid (pre-laying exodus). Colours are consistent within pairs. Most
male birds remain close to the colony while the females move further away, with four moving towards the shelf edge southwest of
Ireland. All females laid an egg immediately after returning to the colony after these trips, except FR88595, which stayed in the
Irish Sea—shown in blue. Bathymetry contours at 1000 m intervals indicate the edge of the continental shelf.
1220 T. Guilford et al. Pelagic seabird movements
Proc. R. Soc. B (2009)
conditions. Sooty shearwaters (Shaffer et al. 2006) can
achieve similar speeds during rapid migration, and Manx
shearwaters were often recorded by GPS moving at
55 km h
K1
or faster during foraging excursions from
their breeding colony (Guilford et al. 2008), presumably
by exploiting wind assistance. Nevertheless, the southern
portion of the return migration does not appear to be driven
primarily by winddynamics, which would predict an easterly
movement from the over-wintering area initially to exploit
the gyre in the south Atlantic (e.g. Felicisimo et al. 2008).
An alternative explanation is that returning migrants
continue to exploit fishing opportunities off the coast of
Brazil as they start their return, which might instead allow
a reduction in the cost of migratory transport by requiring
lighter on-board fat reserves.
It is clear, however, that migration usually takes much
longer than could in principle be achieved in sustained
travel, particularly for males. Closer inspection of the data
reveals that birds do not normally make their journey in a
single flight, but have one or more periods in which there is
little onward progress for up to two weeks at a time. We have
labelled these periods as stopovers, which are equally
common in males and females, and hypothesized that they
may serve the same refuelling function as the traditional
staging areas of terrestrial migrants (e.g. Newton 2008).
Similar staging areas have been noted for black-browed
40°S
20°S
0°
20°N
40°N
60°N
80°W
40°S
20°S
0°
20°N
40°N
60°N
60°W 40°W 20°W 0°80°W 60°W 40°W 20°W 0°
(a)(b)
(c)(d)
Figure 6. The tracks of four birds, two male and two female, with successive points joined by lines (noting that there may be
periods of missing data between two successive points, particularly around the equinoxes). The points are coloured red, green
and blue, with respect to their posterior probability of classification into one of three classes using the supervised classifier. Red,
summer feeding; blue, winter feeding; green, migration. (a) FR53237 male, (b) FB22795 female, (c) FP52700 female,
(d) FC83751 male.
Pelagic seabird movements T. Guilford et al. 1221
Proc. R. Soc. B (2009)
albatrosses (Phillips et al. 2005), but have not so far been
seeninthemostcomparablespecieswhosetrans-
equatorial migrations have been similarly tracked, the
Cory’s (Gonzales-Solis et al. 2007) and sooty (Shaffer et al.
2006) shearwaters, for reasons unknown. A larger dataset
will be required to determine whether there is any
systematic relationship between the timing and position
of stopovers and environmental variables. Nevertheless,
our initial analyses indicate that, while strong headwinds
associated with passing weather systems sometimes
coincide with stopping periods, on other occasions this is
not the case. Hence, while there may be adverse weather
explanations for some stopovers, others cannot be
accounted for in this way, and some clearly outlive the
longevity of weather systems. Stopover sites are recognized
as important areas for terrestrial migrants, but at sea
their distance from land and probably mobile nature may
well have allowed them to go largely unrecognized for
marine migrants.
Apart from a longer return migration in males than
females, there are no obvious differences in movement
patterns between the sexes (although it is clear that
members of a pair do not synchronize migration). The
major exception to this involves the pre-laying exodus
from the colony when we show that for a majority of
females there is a distinct southwest to westerly displace-
ment (near the continental shelf edge) from the normal
breeding-season foraging destinations (see Guilford et al.
2008). Either these destinations must provide females
with special resources for egg formation, or they are simply
rich foraging areas that are out of range for males on nest-
guarding duty and all breeding birds while tending eggs
and chicks. One of the two females not showing this
pattern (FR88595) failed to lay an egg in 2007. Female
black-browed albatrosses also move further from their
colony than do males during their pre-laying exodus
(Phillips et al. 2005).
The three dominant classes of behaviour recognized by
the unsupervised classification of the immersion data
overlap well with those recognized by the supervised
classification trained on expert selected periods of
migration, over-wintering and breeding-period foraging
trips. This allows some confidence in using the unsuper-
vised classes to identify probable activity types at other
stages of the migratory cycle without expert scrutiny or
dependence on location data required by the expert. Our
analysis, which may have general applicability to similar
datasets and which allows the extraction of information
not immediately evident to researchers using informal
human pattern recognition, highlights two striking features.
First, activity at the two core destinations, the Irish Sea
during summer breeding and the Patagonian Shelf during
over-wintering, is distinct. A priori we had no expectation
of this. There is perhaps an obvious hypothesis: that
summer foraging involves periods of movement to and
from the colony, while during winter birds have only to
look after themselves and may thus remain more
sedentary. Scrutiny of the stopover periods in particular,
when there is no return to the colony, indicates a propor-
tion of activity classed most similar to summer feeding,
which argues against the hypothesis that one particular
activity such as commuting determines the classification.
An alternative explanation is that different search or
foraging styles may be required in different areas where
prey type, availability or local distributions themselves
differ (e.g. Weimerskirsch et al. 1994). Differences in wing
loading between winter and summer are another possi-
bility, since shearwaters are thought to moult gradually
over the winter. Periods of active migration, on the other
hand, are distinctively different again, and are clearly
associated with the long-distance migratory movements
between the two core destinations. This can be seen from
the spatial pattern of days classed as migration, and there
is a strong congruence with days where there are large
changes in longitude (e.g. figure 5). But, again, our
unsupervised classification does not require spatial infor-
mation, so it can in principle be used to identify periods of
migration where spatial data are unavailable.
Second, stopover periods, which are defined by spatial
criteria, commonly contain activity patterns more typical
of summer or winter feeding than of active migration
(58% of stopover days overall show winter or summer
foraging type characteristics in equal proportions). This
supports the hypothesis that stopovers may indeed be
refuelling stops integral to the shearwater’s migration
tactics. Such behaviour, which has critical implications
for the protection of important resource areas at sea, has
so far gone little noted in seabirds.
This paper represents a first attempt to use analytical
techniques originally developed within machine learning
to identify behaviour remotely using a combination of
immersion and spatio-temporal data from miniature
geolocation technology. Our fully Bayesian approach is
new because we have initially classified behaviour states
from salt-water immersion data, independently of location
(the closest related work we know of is Roberts et al.
2004). This has the advantage of later allowing cross-
validation with the location data. The aim is not to
produce a comprehensive interpretation of the data or of
Manx shearwater migration: more refined techniques and
larger datasets may be required for this. The aim is to show
how such a cross-disciplinary approach may offer a
potentially powerful and relatively uninvasive way to study
the more elusive life histories of smaller, highly pelagic
seabirds, many of which are of vulnerable status. Our
approachmay be extendable to many other complex datasets
where informal analysis cannot easily tease out important
hidden patterns.
We thank Louise Maurice, Christine Nicol, Tom Evans and
the staff and volunteers of Skomer Island for their field
assistance. The Wildlife Trust for South and West Wales, and
the Countryside Council for Wales, provided logistical or
financial support.
Oxford University Local Ethical Review procedures were
followed throughout this project, which does not fall under
A(SP)A, conformed to ASAB guidelines for ethical research
and was conducted with appropriate BTO ringing licences
and a CCW permit.
REFERENCES
Attias, H. 2000 A variational Bayesian framework for
graphical models. In Proceedings of NIPS 12 (ed. T. Leen
et al.), Cambridge, MA: MIT Press.
Bishop, C. 2006 Pattern recognition and machine learning
New York, NY: Springer. (doi:10.1117/1.2819119)
Brooke, M. 1990 The Manx shearwater. London, UK: T. &
A. D. Poyser.
1222 T. Guilford et al. Pelagic seabird movements
Proc. R. Soc. B (2009)
Felicisimo, A. M., Munoz, J. & Gonzalez-Solis, J. 2008
Ocean surface winds drive dynamics of transoceanic aerial
movements. PLoS One 3, e2928. (doi:10.1371/journal.
pone.0002928)
Fisher, R. A. 1936 The use of multiple measurements in
taxonomic problems. Ann. Eugen. 7, 179–188.
Gonzales-Solis,J.,Croxall,J.P.,Oro,D.&Ruiz,X.2007
Transequatorial migration and mixing in the wintering areas
of a pelagic seabird. Front. Ecol. Environ. 5, 297–301. (doi:10.
1890/1540-9295(2007)5[297:TMAMIT]2.0.CO;2)
Guilford, T. C., Meade, J., Freeman, R., Biro, D., Evans, T.,
Bonadonna, F., Boyle, D., Roberts, S. & Perrins, C. M.
2008 GPS tracking of the foraging movements of Manx
shearwaters Puffinus puffinus breeding on Skomer Island,
Wales. Ibis 150, 462–473. (doi:10.1111/j.1474-919X.
2008.00805.x)
Hamer, K. C. 2003 Puffinus puffinus Manx shearwater. BWP
Update 5, 203–211. (doi:10.1046/j.1365-2435.1997.
00118.x)
Hill, R. D. 1994 Theory of geolocation by light levels. In
Elephant seals: population ecology, behaviour and physiology
(eds B. J. Le Boeuf & R. M. Laws), pp. 227–236. Berkeley,
CA: University of California Press.
Martinez, W. L. & Martinez, A. L. 2008 Computational
statistics handbook with Matlab, 2nd edn. London, UK:
Chapman and Hall.
Newton, I. 2008 The migration ecology of birds. London, UK:
Academic Press. (doi:10.1016/j.tree.2005.04.010)
Pennycuick, C. J. 1969 The mechanics of bird migra-
tion. Ibis 111, 525–556. (doi:10.1111/j.1474-919X.1969.
tb02566.x)
Perrins, C. M. & Brooke, M.de L. 1976 Manx shearwaters in
the Bay of Biscay. Bird Study 23, 295–299.
Phillips, R. A., Silk, J. R. D., Croxall, J. P., Afanasyev, V. &
Briggs, D. R. 2004 Accuracy of geolocation estimates for
flying seabirds. Mar. Ecol. Prog. Ser. 266, 265–272.
(doi:10.3354/meps266265)
Phillips, R. A., Silk, J. R. D., Croxall, J. P., Afanasyev, V. &
Bennett, V. J. 2005 Summer distribution and migration of
nonbreeding albatrosses: individual consistencies and
implications for conservation. Ecology 86, 2386–2396.
(doi:10.1890/04-1885)
Phillips, R. A., Silk, J. R. D., Croxall, J. P. & Afanasyev, V.
2006 Year-round distribution of white-chinned petrels
from South Georgia: relationships with oceanography and
fisheries. Biol. Conserv. 129, 336–347. (doi:10.1016/
j.biocon.2005.10.046)
Phillips, R. A., Croxall, J. P., Silk, J. R. D. & Briggs, D. R.
2008 Foraging ecology of albatrosses and petrels from
South Georgia: two decades of insights from tracking
technologies. Aquat. Conserv. Mar. Freshw. Ecosyst. 17,
S6–S21. (doi:10.1002/aqc.906)
Roberts, S. J. & Penny, W. D. 2002 Variational Bayes for
generalized autoregressive models. IEEE Trans. Signal
Process. 50, 2245–2257. (doi:10.1109/TSP.2002.801921)
Roberts, S., Guilford, T., Rezek, I. & Biro, D. 2004 Positional
entropy during pigeon homing I: application of Bayesian
latent state modelling. J. Theor. Biol. 227, 39–50. (doi:10.
1016/j.jtbi.2003.07.002)
Shaffer, S. A., Tremblay, Y., Awkerman, J. A., Henry, W. R.,
Teo, S. L. H., Anderson, D. J., Croll, D. A., Block, B. A. &
Costa, D. P. 2005 Comparison of light- and SST-based
geolocation with satellite telemetry in free-ranging alba-
trosses. Mar. Biol. 147, 833–843. (doi:10.1007/s00227-
005-1631-8)
Shaffer, S. A. et al. 2006 Migratory shearwaters integrate
oceanic resources across the Pacific Ocean in an endless
summer. Proc. Natl Acad. Sci. USA 103, 12 799–12 802.
(doi:10.1073/pnas.0603715103)
Smith, S., Thompson, G. & Perrins, C. M. 2001 A census of
the Manx shearwater on Skomer, Skokholm and Mid-
dleholm, West Wales. Bird Study 48, 330–340.
Thompson,K.R.2002ManxshearwaterPuffinus puffinus.
In The migration atlas (eds C. Wernham, M. Toms,
J. Marchant, C. Clark, G. Siriwardena & S. Baillie),
pp. 120–123. London, UK: T. & A. D. Poyser.
Weimerskirsch, H., Doncaster, C. P. & Cuenot-Chaillet, F.
1994 Pelagic seabirds and the marine environment:
foraging patterns of wandering albatross in relation to
prey availability and distribution. Proc. R. Soc. B 255,
91–97. (doi:10.1098/rspb.1994.0013)
Weimerskirch, H., Bonadonna, F., Bailleul, F., Mabille, G.,
Dell’Omo, G. & Lipp, H-P. 2002 GPS tracking of
foraging albatrosses. Science 295, 1259. (doi:10.1126/
science.1068034)
Wilson, R. P., Ducamp, J. J., Rees, G., Culik, B. M. &
Niekamp, K. 1992 Estimation of location: global coverage
using light intensity. In Wildlife telemetry: remote monitoring
and tracking of animals (eds I. M. Priede & S. M. Swift),
pp. 131–134. Chichester, UK: Ellis Horward.
Pelagic seabird movements T. Guilford et al. 1223
Proc. R. Soc. B (2009)
... The foraging range of Manx Shearwaters is such that birds in our study could also have been feeding in the waters around Ireland during the breeding season (Wischnewski et al. 2019). However, as Manx Shearwaters winter in South America each year (Guilford et al. 2009), the adults could have ingested the plastics at any point of their journey. Although a higher proportion of fledglings had ingested plastic in their digestive tracts compared to adults, the adults contained a greater number of pieces, with a higher variability in the number of pieces ingested between individuals. ...
Article
Full-text available
Plastic debris is ubiquitous in the marine environment, and seabirds are among the most impacted marine vertebrates.
... Over the last 10-15 years, hundreds of alcids have been tagged with a single geolocator throughout their biogeographic range Reiertsen et al., 2021), but our method does not have the power to identify molt in a sufficient proportion of individuals to robustly investigate population-wide patterns. More complex methods, for instance, using machine learning to identify flightless stopovers (Guilford et al., 2009), usually require large amounts of pre-assigned training data to confidently infer behavior, but may even then be liable to misclassification due to the individual-or colony-level differences in behavior (Bennison et al., 2018). Finerresolution data, such as from accelerometers, would allow us to identify flight with much more confidence (e.g., Patterson et al., 2019). ...
Article
Full-text available
Animal-borne telemetry devices provide essential insights into the life-history strategies of far-ranging species and allow us to understand how they interact with their environment. Many species in the seabird family Alcidae undergo a synchronous molt of all primary flight feathers during the non-breeding season, making them flightless and more susceptible to environmental stressors, including severe storms and prey shortages. However, the timing and location of molt remain largely unknown, with most information coming from studies on birds killed by storms or shot by hunters for food. Using light-level geolocators with saltwater immersion loggers, we develop a method for determining flightless periods in the context of the annual cycle. Four Atlantic puffins (Fratercula arctica) were equipped with geolocator/immersion loggers on each leg to attempt to overcome issues of leg tucking in plumage while sitting on the water, which confounds the interpretation of logger data. Light-level and saltwater immersion time-series data were combined to correct for this issue. This approach was adapted and applied to 40 puffins equipped with the standard practice deployments of geolocators on one leg only. Flightless periods consistent with molt were identified in the dual-equipped birds, whereas molt identification in single-equipped birds was less effective and definitive and should be treated with caution. Within the dual-equipped sample, we present evidence for two flightless molt periods per non-breeding season in two puffins that undertook more extensive migrations (>2000 km) and were flightless for up to 77 days in a single non-breeding season. A biannual flight feather molt is highly unusual among non-passerine birds and may be unique to birds that undergo catastrophic molt, i.e., become flightless when molting. Although our conclusions are based on a small sample, we have established a freely available methodological framework for future investigation of the molt patterns of this and other seabird species.
... Manx shearwaters (Puffinus puffinus), hereafter referred to simply as 'shearwaters', are medium-sized pelagic seabirds of the Family Procellariidae, which circumnavigate the Atlantic Ocean breeding on uninhabited islands in north-west Europe, including those around Britain and Ireland, and overwintering along the South American, principally the Argentinian, coast [1]. Shearwaters spend most of their time at sea during migration and overwintering with limited contact with conspecifics, likely limiting disease spread and contributing to their high annual survival [2,3]. ...
Article
Full-text available
Puffinosis is a disease of a range of seabirds characterised by dorsal and ventral blistering of their webbed feet, conjunctivitis, dry necrosis, leg spasticity, head shaking, loss of balance, tremors, and death. It is associated with Manx shearwaters (Puffinus puffinus), frequently affecting chicks within their underground nesting burrows. The aetiology of the disease is unclear but has been attributed to a type-2 coronavirus associated with Neotombicula mites as a potential vector. However, there is some uncertainty given potential laboratory contamination with mouse hepatitis virus and failure to fulfil Koch’s postulates, with birds injected with isolates remaining healthy. We describe a detailed case report of puffinosis in a Manx Shearwater covering necropsy, histology, bacteriology, and metagenomics including viral sequencing. We found no evidence of viral infection or parasites. Our results are consistent with an entirely environmental aetiology, with caustic faecal ammonia in damp nesting burrows causing conjunctivitis and foot dermatitis breaking the skin, allowing common soil bacteria (i.e., Flavobacterium, Staphylococcus and Serratia spp., Clostridia perfringens and Enterococcus faecalis) to cause opportunistic infection, debilitating the bird and leading to death. A similar condition (foot pad dermatitis or FPD) has been reported in broiler chickens, attributed to caustic faeces, high humidity, and poor environmental conditions during indoor rearing, preventable by adequate ventilation and husbandry. This is consistent with puffinosis being observed in Shearwater nesting burrows situated in tall, dense, vegetation (e.g., bracken Pteridium aquilinum) but rarely reported in burrows situated in well-ventilated, short coastal grasslands. This proposed environmental aetiology accounts for the disease’s non-epizootic prevalence, spatial variation within colonies, and higher frequency in chicks that are restricted to nesting burrows.
... In "Migration and stopover in a small pelagic seabird, the Manx shearwater Puffinus puffinus: insights from machine learning" [2] they used a Bayesian analysis adapted from machine learning techniques to identify distinct behavioural categories inherent in the patterns within the immersion records, and used this to shed light on the birds' behaviour at different stages of the migratory cycle: during summer feeding, winter feeding, migration and egg-laying. In particular, they identify what appear to be migratory stopover periods that we hypothesize may function in the same way as stopovers in terrestrial migrants for refuelling. ...
... TA B L E 1 Summary of the typical differences between capital, intermediate and income migration Manx shearwater, Puffinus puffinus:Guilford et al., 2009; Arctic terns, Sterna paradisaea:Egevang et al., 2010; South Polar skuas, Catharacta maccormicki:Kopp et al., 2011; long-tailed skuas, Stercorarius longocaudus: Sittler et al., 2011) stop more frequently than Cory's shearwaters, suggesting they also rely on mid-journey foraging to fuel migration. ...
Article
Full-text available
Animal migrations represent the regular movements of trillions of individuals. The scale of these movements has inspired human intrigue for millennia and has been intensively studied by biologists. This research has highlighted the diversity of migratory strategies seen across and within migratory taxa: while some migrants temporarily express phenotypes dedicated to travel, others show little or no phenotypic flexibility in association with migration. However, a vocabulary for describing these contrasting solutions to the performance trade‐offs inherent to the highly dynamic lifestyle of migrants (and strategies intermediate between these two extremes) is currently missing. We propose a taxon‐independent organising framework based on energetics, distinguishing between migrants that forage as they travel (income migrants) and those that fuel migration using energy acquired before departure (capital migrants). Not only does our capital:income continuum of migratory energetics account for the variable extent of phenotypic flexibility within and across migrant populations, but it also aligns with theoreticians’ treatment of migration and clarifies how migration impacts other phases of the life‐cycle. As such, it provides a unifying scale and common vacabulary for comparing the migratory strategies of divergent taxa.
... All migratory species modelled here showed increased stranding numbers towards the south, suggesting that the southern area is used more by these species. The southern Brazilian coast and shelf are important feeding grounds for a variety of seabirds, including S. magellanicus (Marques et al., 2018), both Thalassarche and Procellaria species (Colabuono and Vooren, 2007;Bugoni et al., 2009;Dias et al., 2017;Frankish et al., 2020), and P. puffinus (Guilford et al., 2009). Off south Brazil is also where the highest seabird species diversity and abundance of oceanic taxa aggregate (Neves et al., 2006;Daudt, 2019), and where the southern SBB marks the northern limit of the distribution area of many seabirds (Neves et al., 2006;Frankish et al., 2020). ...
Article
Full-text available
We describe the temporal and spatial patterns of three tetrapod groups (sea turtles, seabirds, and cetaceans) stranded in southeast Brazil, based on daily beach surveys of over >800 km of coastline between 2015 and 2020. Patterns were analysed by stranding rates (individuals/1000 km of coastline) of groups and species; for the 14 most numerous species, we also fitted generalized additive models for location, scale, and shape (GAMLSS) using temporal and spatial explanatory variables. We recorded 53850 animals (22738 sea turtles, 28155 birds, and 2957 cetaceans) of 78 species. Higher stranding numbers were observed during winter/spring for all groups, and could be an effect of a higher occurrence/abundance at the region, driven by productive waters, jointly with higher stranding probabilities due to stronger drifting forces. Most modelled species showed temporal and spatial stranding patterns most likely related to their general occurrence/abundance cycles in southeast Brazil, but for Caretta caretta, Chelonia mydas, Pontoporia blainvillei, and Sotalia guianensis local mortality factors may be influencing stranding numbers. By using intensive, wide ranging beach surveys and including survey effort in the analysis, we provide a comprehensive baseline for stranding numbers in the region, and open a pathway for future hypothesis-testing studies to investigate its drivers.
... In contrast to the tailwind preference of nonpelagic birds (2,(5)(6)(7), Procellariiformes typically prefer crosswind or cross-tailwind flight (12,13,15,16). They are also able to fly huge distances (10,(17)(18)(19) without having to offset their aerodynamic drag losses fully through flapping (13,20). Both differences may be the product of a specialized flight mode called dynamic soaring (21), but although this is the only plausible explanation of how albatrosses are able to glide immense distances without flapping (13), dynamic soaring is more challenging to demonstrate in species such as shearwaters that routinely flap their wings during their characteristic rising-falling flight (20). ...
Article
Full-text available
Dynamic soaring harvests energy from a spatiotemporal wind gradient, allowing albatrosses to glide over vast distances. However, its use is challenging to demonstrate empirically and has yet to be confirmed in other seabirds. Here, we investigate how flap-gliding Manx shearwaters optimize their flight for dynamic soaring. We do so by deriving a new metric, the horizontal wind effectiveness, that quantifies how effectively flight harvests energy from a shear layer. We evaluate this metric empirically for fine-scale trajectories reconstructed from bird-borne video data using a simplified flight dynamics model. We find that the birds' undulations are phased with their horizontal turning to optimize energy harvesting. We also assess the opportunity for energy harvesting in long-range, GPS-logged foraging trajectories and find that Manx shearwaters optimize their flight to increase the opportunity for dynamic soaring during favorable wind conditions. Our results show how small-scale dynamic soaring affects large-scale Manx shearwater distribution at sea.
... Interestingly, both geolocation studies on European Storm Petrels showed that the tracked individuals explored different staging areas throughout their non-breeding period (Lago et al. 2019, present study). This is not uncommon in seabirds (Guilford et al. 2009, Egevang et al. 2010, Dias et al. 2012, Seyer et al. 2021 and has already been detected in other storm petrel species (Pollet et al. 2019). Exploring several areas throughout the non-breeding period may preclude the identification and protection of the main areas explored by European Storm Petrels and may expose these individuals to different threats at sea in each of those areas. ...
Article
Full-text available
Determining the non‐breeding distribution and activity patterns of migratory animals is essential to understand the trade‐offs across breeding, moulting and migratory periods and to evaluate the differential levels of exposure of these animals to threats throughout the year. By taking advantage of the current miniaturisation of geolocators, during the breeding season of 2019, we deployed geolocators on ten European Storm Petrels Hydrobates pelagicus breeding on Benidorm Island, Western Mediterranean to determine the distribution and activity patterns of this small seabird species. Of the eight individuals that were recaptured with usable geolocation data, all individuals migrated to the North Atlantic Ocean, ranging between the Canary Islands and south of Iceland. They possibly take advantage of the more productive waters in the North Atlantic areas explored during the non‐breeding period compared to the Western Mediterranean areas explored during the breeding period. This migratory pattern contrasted with the partial migration described for individuals breeding in the Central Mediterranean. Tracked individuals were more synchronous in their prenuptial than in their postnuptial migratory phenology, probably due to a higher probability of breeding success related to earlier arrival to the breeding area. At‐sea activity data indicated that individuals spent more time on the water at night (and especially on those nights around the new moon) during the first half of the non‐breeding period (matching the dates of the moulting period of the species). While birds may begin to moult their feathers while breeding, some seemed to overlap the moulting period with the postnuptial migration. Our results show that multi‐colony studies are needed to understand the drivers of distinct migratory patterns at intra‐ and inter‐population levels and how small migratory species trade‐off the overlap of moult with other energetically demanding activities as breeding or migrating.
Article
In this research, we have studied the dynamics of migration of The Lesser Black-Backed Gull. To study and analyse the data we have used machine learning techniques like Hidden Markov Models. The paper provides a detailed analysis of the migratory patterns of the bird. Through this research we aim to provide a detailed study on the bird and predict its location through machine learning models.
Article
Full-text available
Despite increasing interest in long-distance migration, the wintering areas, migration corridors, and population mix in winter quarters of most pelagic marine predators are unknown. Here, we present the first study tracking migration movements of shearwaters through the non-breeding period. We used geolocators (global location sensing [GLS] units based on ambient light levels) to track 22 Cory’s shearwaters (Calonectris diomedea) breeding in three different areas. Most birds wintered in one or more of three relatively small areas, all clearly associated with major coastal upwelling systems of the tropical and south Atlantic. Trans-equatorial movements were dominated by prevailing trade winds and westerlies, while calm, oligotrophic areas were avoided. Breeding populations clearly differed in their preference amongst the three major wintering areas, but showed substantial mixing. This illustrates the exceptional value of GLS, not only for determining and describing the influence of oceanographic features on migration patterns, but also for assessing population mix in winter quarters. This knowledge is essential to understanding the impacts of population-level threats, such as longlining, offshore windfarms, and oil spills on multiple breeding sites, and will be critical in devising conservation policies that guarantee the sustainable exploitation of the oceans.
Article
Full-text available
Many birds show a surprising degree of intraspecific variability in migratory tendency and choice of wintering site. In this study, we tracked the seasonal movements of 35 nonbreeding Black-browed Albatrosses Thalassarche melanophrys from South Georgia, including 24 birds followed in two consecutive years. This revealed consistent patterns of status-related, sex-specific, and individual variation in wintering strategies, and provided the first description of the summer distribution of failed/deferring breeders. Individuals exhibited a striking degree of site fidelity, returning to the same region (southwest Africa or Australia) and showing correlated centers of distribution, as well as remarkable consistency in the chronology of their movements, in consecutive years. Nonetheless, a degree of behavioral flexibility remained, and particularly on the return migration, birds moved between, or bypassed, alternative intermediate staging sites depending on local circumstances. Initiation of the outward migration varied according to breeding status, timing of failure, and sex: deferring breeders and those that failed early departed two months before successful birds, and successful females departed 1-2 weeks earlier than males. Sex-related latitudinal variation in distribution was also apparent, with females wintering farther north within the Benguela system. Moreover, the only migrant to Australia was a male, supporting an apparent tendency for male-biased breeding dispersal inferred from genetic analyses. Distribution and timing of movements appeared in general to relate to avoidance of competition from congeners and conspecifics from other populations. From a conservation perspective, the study indicated that, for the declining Black-browed Albatross population at South Georgia, the primary focus should be toward improving the management (especially reducing bycatch levels) of fisheries in the central and eastern South Atlantic.
Article
For this account, the sections in BWP Vol. 1 on Field characters, Habit, Distribution, Population, Movements, Food, Social pattern and behaviour, Voice, Breeding, Plumages, Bare parts, Moult, Measurements, Weights and Structure have been updated or rewritten and new sections added on Survival and age structure, Conservation, Breeding performance, and Taxonomy and phylogeny. In particular, new data have been added from studies of vocal behaviour and breeding and foraging ecology, and from studies of shearwater phylogeny and systematics. World population 0.4-0.5 million pairs; largest colonies on islands in SW Wales and on Rum, West Scotland. Breeding distribution also extended to North America since mid-1970s. Annual survival high (c. 90%) and oldest bird to date at least 52 years. Fledglings at Welsh colonies affected by puffinosis, but kills <5% of fledglings even in a severe year and has no significant impact on population size. Very vulnerable, however, to introductions of exotic predators; e.g. now close to extinction on Canna, Scotland following introduction of brown rats. Voice of males and females distinct; highly important in sexual display and defence of breeding burrows. Food mainly small clupeiform fish and squid. Chicks in SW Wales fed every 1.2 nights on average; >40% more food provided by male parents than by females, as a result of shorter foraging trips (male 1-4 nights, female 1-7 nights). Laying date advances with age up to c. 10 years, and egg size increases up to onset of senescence (c. 16 years), but breeding performance probably related more to pair bond than to age or experience. Mediterranean-breeding shearwaters treated as subspecies of Manx Shearwater in BWP Vol. 1, but now widely regarded as separate species.
Article
This book presents an up-to-date, detailed and thorough review of the most fascinating ecological findings of bird migration. It deals with all aspects of this absorbing subject, including the problems of navigation and vagrancy, the timing and physiological control of migration, the factors that limit their populations, and more. Author, Ian Newton, reveals the extraordinary adaptability of birds to the variable and changing conditions across the globe, including current climate change. This adventurous book places emphasis on ecological aspects, which have received only scant attention in previous publications. Overall, the book provides the most thorough and in-depth appraisal of current information available, with abundant tables, maps and diagrams, and many new insights. Written in a clear and readable style, this book appeals not only to migration researchers in the field and Ornithologists, but to anyone with an interest in this fascinating subject. * Hot ecological aspects include: various types of bird movements, including dispersal and nomadism, and how they relate to food supplies and other external conditions * Contains numerous tables, maps and diagrams, a glossary, and a bibliography of more than 2,700 references * Written by an active researcher with a distinguished career in avian ecology, including migration research.
Article
The foraging success of pelagic seabirds and where and when they catch prey has been largely unknown until now. We use satellite transmitters in conjunction with recorders measuring feeding times and masses ingested to show that wandering albatrosses (Diomedea exulans) on foraging trips from the nest encountered prey on average every 4.4 h and consumed 2.1 kg of food daily. They travelled up to 3600 km from the nesting colony in search of scarce prey, mostly pelagic squid. These were distributed without relation to underwater topography, and were principally caught during daylight hours in discrete patches widely separated along the foraging route. When foraging inshore, the birds fed on more localized aggregations of squid and fish on the shelf breaks. These results demonstrate that the foraging patterns of large seabirds such as wandering albatrosses, which rely on scarce prey, can help to increase our knowledge of the distribution and availability of poorly known but widespread species of squid.