How and why Aldabra giant tortoises hunt birds and other animals

Full text

Testudo Vol. 9 No. 5 1
How and why Aldabra giant tortoises hunt
birds and other animals
Georgia Miller1, Mia Messenger1, Anna Zora2 and Justin Gerlach1*
1Peterhouse, Trumpington Street, Cambridge CB2 1RD, U.K.
2Frégate Island Foundation, Frégate, Seychelles
*Corresponding author: Justin Gerlach, email: jg353@cam.ac.uk
Introduction
In 2021 we published a description of an Aldabra giant tortoise (Aldabrachelys
gigantea) catching and eating a bird on Frégate Island, Seychelles, providing
the first video evidence of this behaviour (Zora & Gerlach 2021). Prior to
this, there were anecdotal reports of giant tortoises killing birds but these
had not been properly documented. Since our publication a number of other
observations of tortoises eating meat, and in a few cases catching animals,
were reported to us. It is also well known that most tortoise species will
opportunistically eat meat (e.g. when encountering carrion). However, the
not infrequent sightings of predation on sea-birds on Frégate Island remains
distinctive in having been observed on many occasions over several years, and
involving many different individual tortoises (Zora pers. obs.).
The significance of the behaviour cannot be determined from the existing
published evidence: it may be an occasional, opportunistic activity by a small
number of individuals or a more systematic occurrence. It may be a new
phenomenon, although it has been speculated to be a formerly normal part
of giant tortoise and sea-bird interactions which had not been possible on
most islands until habitat and species restoration became advanced (Zora
& Gerlach 2021). Other questions can also be raised: how widespread is
it in the tortoise population? Is it evenly distributed demographically, or is
it more frequent in individuals with higher protein demands (e.g. juveniles
and breeding females)? How often do individual tortoises hunt? What is the
success rate and how important is it for them? What role do other species of
animal play in the diet?
Following the publication of the video observation in 2021 we have started
investigating the behaviour systematically. Here we report on the first results
of these further studies.
Study Site
Frégate island covers 219 hectares (Fig. 1). Naturally it was a wooded island
but was largely converted to agricultural plantation in the 19th century. It
is now managed for conservation and ecotourism and 30 years of habitat
restoration have regenerated woodland over much of the island. Tortoises
© British Chelonia Group + Justin Gerlach 2023

2 Testudo Vol. 9 No. 5
were reintroduced to the island in the 1940s and some 3,000 animals are
now present (Gerlach et al. 2013). Sea birds have recolonised, with the main
tern breeding colony being in the Anse Parc area. There they nest in Pisonia
grandis trees during June to September. High winds in this season frequently
result in chicks being blown out of their nests. When on the ground these
are rarely tended by their parents and fall victim to the abundant predators:
snakes, lizards and crabs.
Methods
Tortoise behaviour on Frégate Island was studied by two of the authors
(GM, MM, with support and assistance from AZ) in July-August 2022. Four
approaches were taken: camera traps, supplementary observation of focal
tortoises and of fallen chicks, and faecal analysis. The main focus of research
was on the tortoises within the lesser noddy (Anous tenuirostris) breeding
colony above the Anse Parc area, where most predation events had been
observed previously. Additional observations were also made elsewhere on
the island, although not systematically.
Tortoises recorded interacting with birds were sexed and assigned an
approximate age category based on size. Sexing was by means of tail length
(when held to the side female tails reach less than half-way from mid-line
to the base of the hind leg, whereas male tails make contact with the base
of the hind leg) and age categories were: juvenile (approximately 30cm
Fig. 1. Simplified habitat map of Frégate Island showing main study locations and the localities
where tortoises were found consuming feathers or birds (observation and faecal analysis) or
were recorded attacking birds.
Bare rock
Beach
Wetland
Hotel
Grass
Scrub
Bamboo
Open partially planted with native trees
Banyan trees
Coconut dominated woodland
Woodland
feathers or birds eaten
attacks on birds
start of Grande Anse Road
Anse Parc
Airstrip
Pirate ruins
plateau
Grande Anse
200 m
© British Chelonia Group + Justin Gerlach 2023

Testudo Vol. 9 No. 5 3
straight carapace length), subadult (30-40cm) and adult (over 40cm). Terns
were categorised as adult (full flying), large fledglings (able to flutter off the
ground or to fly short distances), medium (at least half full size but flightless)
and small (less than half full size, lacking any flight feathers). All terns
observed in the interactions were lesser noddies, although white terns (Gygis
alba) were also present.
Camera traps
Four camera traps were initially set up in areas that contained nesting sites
of the lesser noddy, principally in habitats classed as mixed woodland and
open partially planted with native trees (Fig. 1). Initially the cameras were
positioned to capture at least one lesser noddy chick that was on the ground
or otherwise in reach of giant tortoises. Where possible, sites were selected
where tortoises were active nearby to maximise the chance of recording an
encounter. The camera traps were then left to run for approximately 20 hours
before their SD cards were switched. If the original chick was no longer in
view at this point the camera was repositioned. Camera traps were set up
to record throughout the day and night, recording 30-60 seconds of video
following motion detection, with a 60 second interval.
During the study two of the camera traps stopped working due to battery
faults. This limited the areas that could be monitored simultaneously. After
approximately a week, camera position was varied on a daily basis to cover
more regions, still within the area of the lesser noddy colony. When no fallen
chicks were observed on the floor, camera traps were instead positioned to
cover a wider area to increase the chance of picking up fallen chicks and/or
tortoise predatory behaviour. This included exotic scrub with banyan trees
and native woodland habitats. Once tortoise-bird interactions were observed
in an area, camera traps were placed there more frequently. The aim of
this was to maximise the number of recorded observations, rather than to
quantify the frequency of occurrences.
The footage from the SD cards was reviewed daily between 12:00 and
16:00 and clips with evidence of interactions were kept, and the nature of
this interaction recorded. Interactions were separated into giant tortoises
approaching chicks (walking directly towards a chick with the neck
outstretched, but mouth not open), attacking (approaching with the mouth
open, attempting to bite the chick), killing (live chick at start of observation
but dead by the end) and eating (eating a dead bird, or with feathers in its
mouth).
© British Chelonia Group + Justin Gerlach 2023

4 Testudo Vol. 9 No. 5
Observations
Separate from camera trap observations, we noted down any interactions
we saw across the island. This included the previously stated categories of
observations as well as giant tortoises already in the process of eating birds
– it is therefore unknown whether these had been hunted or were already
dead and picked up from the ground. Other staff working on the island also
noted down instances of tortoises in the process of eating birds, recording
location, time and where possible the sex of the tortoise.
In addition to these casual recordings, focal observations were carried
out on 50 occasions between 09:00 and 11:30 by locating one or more
lesser noddy chicks on the ground and observing their behaviour and the
behaviour of any nearby tortoises. Notable interactions here were recorded
in both note form, and also via the use of mobile phones to record photos
and videos of the interactions.
Fallen chicks
Chicks that had fallen from their nests and were on the ground or
otherwise potentially in reach of a tortoise (on top of rocks, branches and
tree roots) were noted. The time, location, habitat, chick position, and
species were recorded. We estimated the size of the chick visually (small,
medium, large) and the distance between the chick and nearest giant
tortoise in metres, as well as the presence of other nearby tortoises less
than 10m from the chick. The behaviour of both the chick and tortoise was
described.
Faecal data
Faecal data was collected by dissecting individual tortoise faeces using sticks
and tweezers to evaluate the frequency of consumption of birds. A visual
estimate of the percentages of the main constituent items was made. These
items were categorised as leaves, grass, seeds, giant millipede rings and
bird remains, including single feathers. Seeds comprised large fruit seed;
grass seeds were omitted from this category. Unidentifiable material was
categorised as ‘other’. Faeces were examined in 11 different locations, in the
analysis these are grouped into 8 main areas.
Results
Camera traps and observations
Camera traps recorded 23 interactions, of which 22 were aggressive
approaches by tortoises or tortoises eating birds (Fig. 2). The remaining
observation was a chick standing on the back of a tortoise for over 4 hours.
Of the aggressive interactions 11 were approaches, 3 attacks, 2 kills and
6 eating. Casual observations recorded 10 interactions: 4 approaches,
© British Chelonia Group + Justin Gerlach 2023

Testudo Vol. 9 No. 5 5
2 definite attacks and 1 kill, in addition to 3 tortoises eating birds. The
proportions of these behaviours observed at different times and categorised
by sex are shown in Fig. 3.
Fallen chicks
Observations of 143 fallen chicks (6 white terns, 137 lesser noddies, including
3 dead) recorded no interactions. In addition camera traps caught two cases
of tortoises ignoring dead chicks (both near hydroponics) and 5 cases of
tortoises ignoring live chicks within 3m (above Anse Parc, Anse Parc road,
plateau) and 10 within 1m (plateau, above Anse Parc, Pirate ruins, Airstrip,
clearing of tortoise trail).
Faecal data
A total of 377 faeces were dissected from 8 main areas. In all cases leaves,
grass and fruit predominated, comprising at least 95% of the faeces from
all areas (Fig. 4). Excluding plant matter, the main animal component was
found to be giant millipedes Sechelleptus seychellarum. Bird remains were
recorded in only 5 localities, making up no more than 19% of non-plant
diet (Fig. 4). 38% of samples with feathers contained a single feather, small
numbers (up to 5) were found in a further 10 samples. Larger numbers
were found in samples from the Pirate ruins (20 feathers and additional
uncounted clumps of feathers) and plateau (3 feathers with an entire bird’s
head (Fig. 5)).
Fig. 2. Examples of interactions between tortoises and lesser noddies: (A) chick riding on
tortoise; (B) tortoise approaching a chick; (C) attacking; (D) eating
A B
C D
© British Chelonia Group + Justin Gerlach 2023

6 Testudo Vol. 9 No. 5
Fig. 3. Aggressive interactions between tortoises and birds at different times
of day, comparing data from camera traps and observations categorised by (A)
tortoise sex and (B) type of interaction, showing the peak of activity around
09:00-10:00 and the predominance of attacks by female tortoises.
A
B
06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00
06:00 07:00 08:00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00
© British Chelonia Group + Justin Gerlach 2023

Testudo Vol. 9 No. 5 7
Discussion
Although the present study recorded a relatively small number of attacks by
tortoises on birds (23 plus 10 records of tortoises eating birds) the interactions
were sufficient to reveal some patterns to the behaviour. Hunts (including
approaches, attacks and kills) were observed in five areas (plateau, Pirate
ruins, above Anse Parc, tortoise trail clearing, start of Grande Anse road) and
throughout the day, but with frequencies that depart from general activity
patterns, showing a strongly bimodal pattern in the morning (06:00 and
09:30) but little afternoon activity. In contrast general activity recorded in this
species on other islands has a main peak at 08:00 and a smaller one at 18:00
(Falcón et al. 2018). This may suggest that dawn (before the general rise in
activity) represents a particularly attractive hunting time.
Fig. 4. Faecal analysis showing the percentage of faecal samples containing different items.
Top: Animal material. Below: All material recorded.
Locations: 1: Airstrip (n=41), 2: near Airstrip (36), 3: start of Grande Anse road (20),
4: Anse Parc (139), 5: Pirate ruins (31), 6: tortoise trail near Anse Parc (21), 7: plateau (64),
8: Grande Anse trail (25)
© British Chelonia Group + Justin Gerlach 2023

8 Testudo Vol. 9 No. 5
Hunts (excluding eating, which may represent carrion feeding) were
observed most on the plateau (4: 2 approach, 1 attack, 1 kill), with a single
observation at the Pirate ruins (attack), above Anse Parc (approach) and
tortoise trail clearing (approach). Camera traps recorded more hunts at
the Pirate ruins (8: 4 approach, 3 attack, 1 kill) than on the plateau (6: 5
approach, 1 kill). The differences in the two survey methods probably reflect
the differences in survey effort. Differences between the Pirate ruins and the
plateau are impossible to evaluate as data collection was not standardised.
Camera traps also recorded single approaches in the tortoise trail clearing
and at the start of the Grande Anse road. In the behaviours recorded, casual
observations and camera trapping seem to have been broadly comparable;
casual observations recorded a slightly lower frequency of approach
(0.4 compared to 0.5) and a slightly higher attack rate (0.2 compared to
0.13). This may suggest that the casual observations missed some of the
approaches, which may be due to the majority of approaches recorded by
camera traps being at times of day when observations were scarce.
Of hunts recorded to their end, 13% were successful (13% on the camera
traps, n=16; 14% in casual observation, n=7). This is a high proportion and is
in marked contrast to the large number of observations of tortoises ignoring
chicks. This may suggest that hunting only occurs when tortoises have a high
confidence of success, or a high level of motivation. The duration of hunts varied
considerably, from 2 to at least 18 seconds, observation suggests that this was at
least in part due to the ease or otherwise of moving over the rocky ground.
Attacks (excluding observations of just eating) were targeted at large
fledglings (8) and medium chicks (8), with fewer attacks on adults (5) and
small chicks (2). The highest successful kill rates were of fledglings (25%)
Fig. 5. Bird’s head (two views) and fish skeleton found in tortoise faeces
© British Chelonia Group + Justin Gerlach 2023

Testudo Vol. 9 No. 5 9
and medium chicks (14%). The camera recordings show that the majority of
medium and small chicks were able to run away from the tortoises and were
rarely pursued for more than a few steps. Despite having at least some flight,
fledglings seemed to be more vulnerable to attack as they let the tortoises
come closer than did smaller chicks and when taking avoiding action their
fluttering wings made a large target for the tortoises. In all captures that
could be seen clearly the tortoise first caught the bird by a wing-tip. One
observation of two tortoises eating a dead chick is notable as the second
tortoise had walked past the chick earlier and only approached once the
first tortoise had started eating, this may suggest that they are not detecting
carrion by scent.
Faecal analysis indicates that non-plant matter is a small component
of the diet, but will be biased towards the least digestible components:
cellulose in plant tissue and chitin and keratin in animal matter. Bird remains
were scarce (up to 19% of non-plant diet in only 5 areas), and were most
abundant on the plateau and Pirate ruins. Unexpectedly millipedes were
present in faeces from all areas, forming 1-2% of the diet. They formed a
substantial proportion (15-25%) of nine faecal samples from the Grande
Anse trail (1), Pirate ruins (2), Anse Parc (3) and the plateau (3). This is a
completely new addition to the food list for the Aldabra giant tortoise
although it is known that live millipedes are consumed normally by Kinixys
tortoises in Africa (Hailey et al. 2001). The Seychelles giant millipede is
abundant on Frégate Island and potentially a significant source of protein
for the tortoises. They do secrete defensive benzoquinones which would
be expected to be a deterrent, however, not all individuals do this (Gerlach
pers. obs.) and it is possible that the tortoises consume these more palatable
animals. Alternatively, they may be consumed as carrion or accidentally in
leaves, although observations suggest that at least some are consumed
alive (Zora pers. obs.). Their presence as a notable dietary component is a
significant new finding and needs to be examined in much greater detail in
the future.
It is striking that the majority of interactions were made by female tortoises
(0.82 of observations, 0.68 of camera traps). Excluding unsexed individuals
the proportions are 0.88 female (0.9 observations, 0.84 camera), 0.12 male
(observations 0.1, camera 0.17) (n=33). The sex ratio of adult tortoises on
the island has been estimated to be slightly male biased at 45:55 (Gerlach et
al. 2013), so these data indicate that hunting behaviour shows a strong sex
bias, although it is not exclusively pursued by females.
Despite the frequency of bird hunting by giant tortoises on Frégate, they
cannot be regarded as a threat to the terns as they are only catching flightless
chicks on the forest floor. These chicks have almost no prospect of survival
on the ground, facing predation from snakes, skinks and crabs.
© British Chelonia Group + Justin Gerlach 2023

10 Testudo Vol. 9 No. 5
The present study has expanded our base of information on this remarkable
behaviour involving the hunting of birds, but also on the role that other
animal matter may play in the natural diet of Aldabra tortoises. Although
it is too early to evaluate its full significance it is now apparent that several
giant tortoises on Frégate Island actively hunt birds and that it is a behaviour
pursued predominantly by adult female tortoises. Further animal protein is
consumed in the form of giant millipedes. Whether millipede consumption
is also a female associated behaviour is at present unknown. Further studies
will endeavour to quantify the rate of both hunting and carrion consumption
in order to evaluate how important this behaviour is to individual tortoises.
Acknowledgements
We are grateful to Frégate Island Foundation for supporting this work and
to the island staff for assistance in the field. GM and MM were supported by
travel grants from Peterhouse and equipment for a supplementary study was
provided by the British Chelonia Group. This research was carried out under
permit from the Seychelles Bureau of Standards.
References
Falcón, W., Baxter, R.P., Furrer, S., Bauert, M., Hatt, J.M., Schaepman-Strub, G.,
Ozgul, A. et al. (2018). Patterns of activity and body temperature of Aldabra
giant tortoises in relation to environmental temperature. Ecology and Evolution
8(4): 2108-2121.
Gerlach, J., Rocamora, G., Gane, J., Jolliffe, K. & Vanherck, L. (2013). Giant tortoise
distribution and abundance in the Seychelles islands: past, present and future.
Chelonian Conservation and Biology 12(1): 70-83.
Hailey, A., Coulson, I. M., & Mwabvu, T. (2001). Invertebrate prey and predatory
behaviour of the omnivorous African tortoise Kinixys spekii. African Journal of
Ecology 39(1): 10-17.
Zora, A. & Gerlach, J. (2021). Giant tortoises hunt and consume birds. Current
Biology 31(16): R989-R990.
© British Chelonia Group + Justin Gerlach 2023