Behavioral
Ecology
The official journal of the
ISBE
International Society for Behavioral Ecology
Behavioral Ecology (2014), 25(5), 1268–1275. doi:10.1093/beheco/aru122
Original Article
Optimal hunting conditions drive circalunar
behavior of a diurnal carnivore
Femke Broekhuis,a,b Steffen Grünewälder,c,d John W. McNutt,b and David W. Macdonalda
aDepartment of Zoology, Wildlife Conservation Research Unit, University of Oxford, Recanati-Kaplan Centre,
Tubney House Abingdon Road, Tubney, Oxfordshire OX13 5QL, UK, bBotswana Predator Conservation Trust,
Private Bag 13, Maun, Botswana, cCentre for Computational Statistics and Machine Learning, Graham
Knight Computer Science Department, University College London, Gower Street, London WC1E 6BT, UK,
and dDepartment of Computer Science, University College London, Gower Street, London WC1E 6BT, UK
Received 2 January 2014; revised 13 June 2014; accepted 15 June 2014; Advance Access publication 5 August 2014.
Foraging requirements and predation risk shape activity patterns and temporal behavior patterns widely across taxa. Although this
has been extensively studied in small mammals, the influence of predation and prey acquisition on the activity and behavior of large
carnivores has received little attention. The diurnal activity described as typical for cheetahs (Acinonyx jubatus) has been explained in
terms of their avoidance of antagonistic interactions with other larger predators. However, a recent study revealed that cheetahs are
frequently active at night, especially during periods of full moon. Being both predator and “prey” in an environment with comparatively
high densities of larger and competitively dominant nocturnal predator species, we investigated whether cheetah nocturnal behavior could be explained by favorable conditions for 1) predator avoidance or 2) prey acquisition. We used a data set of continuously
recorded behavior created using machine-learning techniques on behavioral data collected in the field to transform recorded 2D activity values from radio-collars into 3 distinct behavioral states (feeding, moving, and resting). We found that 32.5% of cheetah feeding
behavior occurred at night and that, in the dry season, nocturnal feeding behavior was positively correlated with moonlight intensity.
Our results suggest that nocturnal and circalunar behavior of cheetahs is driven by optimal hunting conditions, outweighing the risks
of encountering other predators. Using novel methodology, the results provide new insights into the temporal distribution of behavior,
contributing to our understanding of the importance of moonlight and season on the behavior patterns of diurnal species.
Key words: circalunar behavior, diurnal carnivore, machine-learning, moonlight, optimal hunting, time budgets.
Introduction
Foraging requirements and predation risk shape activity patterns
and temporal patterns of behavior widely across taxa (Schoener
1974; Kronfeld-Schor and Dayan 2003). For example, Fenn and
Macdonald (1995) showed that nocturnal Norwegian rats (Rattus
norvegicus) shifted to diurnal activity to escape predation by a nocturnal predator (Vulpes vulpes), and Harrington et al. (2009) showed
a similar shift by mink (Neovison vison) in response to the presence
of competitively dominant otters (Lutra lutra) and polecats (Mustela
putorius). A complete switch in activity is rare however as different levels of light availability require anatomical, physiological,
and behavioral adaptations (Daan and Aschoff 1975; KronfeldSchor and Dayan 2003). As a result, patterns of activity are often
entrained by day–night cycles, underpinned by hormone physiology, which leads to species either being active during the day (diurnal) or active at night (nocturnal; Daan and Aschoff 1975; Klein
Address correspondence to F. Broekhuis. E-mail: [email protected].
© The Author 2014. Published by Oxford University Press on behalf of
the International Society for Behavioral Ecology. All rights reserved. For
permissions, please e-mail: [email protected]
et al. 1997). Despite this, activity patterns are not completely rigid
and subtle changes to temporal patterns of animal behavior are
expected to reflect adaptive, if not optimal, cost–benefit responses
to variation in both resource acquisition and the risk of competition or predation (Daan and Aschoff 1975; Lima and Bednekoff
1999; Kronfeld-Schor and Dayan 2003; Kotler et al. 2010). In a
recent study, Prugh and Golden (2014) found that nocturnal species
that depend primarily on vision are more likely to be active during
periods of increased moonlight because it increases foraging efficiency and the ability to detect predators. Numerous studies have
focused on the influence of moonlight on the behavior, activity,
and interactions of nocturnal species (e.g., Brown and Kotler 2004;
Griffin et al. 2005; Penteriani et al. 2013; Prugh and Golden 2014).
However, the role of moonlight on the behavior of diurnal species,
particularly those that are both predator and “prey”, is still poorly
understood. Similar to the circalunar behavior of nocturnal species, diurnal species that rely on vision could likewise adjust their
nocturnal activity to optimize predator avoidance, increase foraging efficiency, or both.
Broekhuis et al. • Circalunar behavior of a diurnal carnivore
Cheetah (Acinonyx jubatus) are predominantly diurnal, a trait that
is rare among the Felidae (Gittleman 1986). It is believed that this
is an evolutionary adaptation to avoid risks associated with larger,
more competitive nocturnal carnivores, in particular lions (Panthera
leo) and spotted hyaenas (Crocuta crocuta; Durant 1998; Hayward and
Slotow 2009). These 3 species compete over similar prey resources
(Mills and Biggs 1993; Hayward and Kerley 2008) but cheetah have
a competitive disadvantage in that they are morphologically smaller
and predominantly solitary (Caro 1994). Cheetahs are therefore not
only a predator but also “prey” as they fall victim to lions and spotted hyaenas through predation and kleptoparasitism (Laurenson
1994; Hunter et al. 2007b). Despite this, a recent study has shown
cheetahs are more active at night than previously described and
that nocturnal activity correlates with moonlight (Cozzi et al. 2012).
This raises an interesting question: Why would this diurnal species
be active at night, particularly at times of increased moonlight?
The nocturnal activity patterns of cheetahs could simply reflect
the increased ability to detect and avoid equally active predators as
previous studies have shown that cheetahs move away from lions and
spotted hyaenas when in close proximity (Durant 2000; Broekhuis
et al. 2013). Alternatively, Cozzi et al. (2012) hypothesized that this
circalunar activity may be explained by benefits in terms of prey
capture: that with increased moonlight, hunting success increases,
offsetting risks of encountering lions or spotted hyaenas. However,
behavioral data at night are difficult to come by, especially for elusive and wide-ranging carnivores, and Cozzi et al. (2012) were therefore unable to distinguish between these 2 hypotheses.
We circumvent these difficulties by using a novel machine learning technique to transform the activity data set used by Cozzi et al.
(2012) into distinct behavioral states: resting, moving, and feeding.
This data set was obtained using direct behavioral observations
to quantify and transform continuous cumulative activity scores
recorded by data-loggers fitted on cheetahs. This enabled reliable
categorization of cheetah behavior at 5-min intervals resulting in a
unique, continuous, and fine-scale behavioral data set. Importantly,
it allowed for behavioral data collection during times when cheetahs could not be directly monitored. In addition, data on several
individuals could be collected simultaneously over a relatively long
period allowing for the investigation of longer term patterns such
as those influenced by the lunar cycle and season. Using these data,
we investigated the diel composition of feeding, moving, and resting behavior of cheetahs.
We hypothesize that if the nocturnal activity described by Cozzi
et al. (2012) is a response to interactions with predators and competitors, cheetah would not feed at night because this is when they
are most vulnerable to antagonistic encounters with their nocturnal
counterparts (Hunter et al. 2007b). However, if this nocturnal activity is related to optimal hunting conditions, then we would expect a
relative increase in hunting success as evidenced by nocturnal feeding behavior. Being visual hunters, we would expect cheetah nocturnal feeding behavior to be positively correlated with moonlight
intensity. If this is the case, then we would expect there to be more
feeding events during the dry season when cloud cover does not
obscure moonlight. Behavioral adaptation to risk and hunting conditions can be investigated even further. If all else is equal, then we
would expect no detectable differences in activity characteristics of
feeding bouts (e.g., frequency and duration) irrespective of time of
day or presence of moonlight.
Continuous and contemporaneous data sets of cheetah behavior
in this study population provide a unique opportunity to determine
the adaptive value of nocturnal behavior in a diurnal species.
1269
Methods
Study area
The study was conducted in the Okavango Delta, a permanent
inland delta in Northern Botswana. The core study site (19°31′S,
23°37′E; elevation ca. 950 m) encompasses an area of approximately 2720 km2, including the Southern part of the Moremi
Game Reserve and the adjacent Wildlife Management Areas (for
details, see McNutt 1996). The area is characterized by 2 distinct
seasons: a cool, dry winter (April to October) and a hot, wet summer (November to March) with a mean annual rainfall of 450–
600 mm/year (Mendelson et al. 2010). Daylight varies from the
shortest day to the longest day by 2 h and 15 min (05h40–19h15
local time on 21 June, to 04h30–20h20 21 December; http://
aa.usno.navy.mil/data/).
Cheetah data
Between October 2008 and July 2011, Global Positioning System
(GPS) radio-collars (VECTRONIC Aerospace GmbH, Germany)
were fitted on all known adult cheetahs in the study area (n = 6).
The GPS radio-collars were equipped with 2 single-axis accelerometers providing quantitative activity data for both the lateral (y axis)
and the anterior–posterior (x axis) movement. The animals’ acceleration on these 2 axes was continuously recorded 4 times/s and
cumulative scores were logged every 5 min for the duration of the
study. Collars recorded activity continuously for a mean of 332 days
(range: 282–381 d). During the time when the collars were active,
each cheetah was found at least once every 2 weeks to collect behavioral data. Behavioral data were then recorded continuously for a
minimum of 1 h, resulting in 188 h of direct observations. During
each observation, 3 fundamental behavior categories were recorded:
feeding, moving, and resting. Each time the focal animal changed
its behavioral state, the exact time was recorded using a handheld
GPS device (Garmin eTrex HC; Garmin, Olathe, KS). This timestamp allowed for the precise synchronization with the activity data
recorded by the collars (see Data processing for more details). To
minimize the impact on cheetah behavior, cheetahs were followed
at a minimum distance of 30 m and observations were conducted
during twilight and daylight hours so that the cheetahs would not be
disturbed by the necessary use of artificial lighting.
In compliance with Botswana law, all immobilizations for
deployment/removal of radio-collars were performed by a
Botswana registered veterinarian. Cheetahs were free-darted and
immobilized using a combination of ketamine (2–2.5 mg/kg) and
medetomidine (0.07 mg/kg), remotely administered by a DanInject CO2 rifle (Dan-Inject, Denmark), and reversed with atipamezol (0.3 mg/mL; following Kock et al. 2006). Sedation time was
kept to a minimum, typically less than 1 h. After immobilization,
all cheetahs recovered fully, showing no signs of distress, and no
apparent side effects were observed on both the short and long
term. Radio-collars were fitted on adults only and weighed 300 g,
0.67 ± 0.10% (mean ± standard deviation [SD]) of an adult’s
body weight (mean ± SD = 45.5 ± 7.0 kg; range = 38–57 kg). On
completion of the study, all radio-collars were removed. The animal handling protocols used conformed to the standards of the
American Society of Mammalogists (Sikes and Gannon 2011) and
were approved by both the Zoology Ethical Review Committee, a
subsidiary of Oxford Universities Animal Care and Ethical Review
(ACER) Committee (License CER-FB2008) and by the Botswana
Department of Wildlife and National Parks (permit EWT 8/36/4).
Behavioral Ecology
1270
Data processing
To investigate patterns of behavior over several moon cycles and
seasons, a large number of data points are required, something that
is not always achievable with purely observational data. We therefore used a novel technique enabling us to gather large amounts
of behavioral data over a relatively short space of time by transforming the activity data from the GPS radio-collars into distinct
behavioral states.
Based on the time-stamp, behavioral observations recorded in the
field were used to label the activity data collected by the activity sensors in the GPS collars. This labeled data set was then used to classify
the remaining, unlabelled, activity data by means of machine-learning techniques (Bishop 2008). The technique we used was a combined Support Vector Machine (SVM) and Hidden Markov Model
(HMM) approach. The SVM is a common method for classification
(Shawe-Taylor and Cristianini 2004), splitting the activity data from
the collars into 3 different behavioral categories: resting, moving,
and feeding. However, because the classifier does not take the temporal sequence of the data into account, we smoothed the classification using a HMM approach, thereby minimizing misclassifications.
This trained classifier was then applied to each of the 5-min activity
values resulting in a full sequence of predicted cheetah behavior for
the entire time that the collars were operational (see Grünewälder
et al. 2012 for full details). The performance of the algorithm was
evaluated using a leave-one-out cross validation approach (see
Table 1 for results). This is a standard approach for performance
evaluation and works in the following way: the SVM was trained on
all but one data set (defined as a block of continuous observations),
and the performance of this learned SVM was then measured on
the omitted data set. This process was then repeated for each of the
data sets and the performance was averaged over all data sets. The
behavior categories were characterized by specific activity values
(mean ± SD—Resting: X = 5.78 ± 2.10, Y = 4.28 ± 1.94; Mobile:
X = 117.77 ± 16.37, Y = 116.87 ± 25.67; Feeding: X = 74.98 ± 26.40,
Y = 53.68 ± 17.77). Because of the similarity between activity values
for feeding and mobile behavior, some misclassification occurred.
However, misclassified behaviors are likely randomly distributed
over the entire data set (i.e., not biased toward time of day, season,
or moonlight) and therefore should not affect the overall results.
This predictive data set was used for all subsequent analyses using
a 2-stage process to investigate the temporal distribution of cheetah
behavior. First, the entire data set was used to investigate the proportion of time cheetahs spent in each behavioral state. The second part of the analysis focused on the feeding behavior of cheetahs
because this is when cheetahs are least vigilant and most at risk of
kleptoparasitism and predation by lions and spotted hyaenas (Lima
1987; Hunter et al. 2007b). Feeding bouts consisted of one or more
feeding states, defined as a continuous block of feeding data points
not separated by mobile behavior (Figure 1). The proportion of time
cheetahs spent feeding is a function of both the feeding frequency
and the length of feeding bouts. Both these components were analyzed in relation to the diel phase (day vs. night), the seasons, and
the proportion of visible moon (which is a proxy for the amount of
moonlight). Only feeding bouts that took place exclusively during
the day or night were considered for further analysis.
Predictor variables
To determine whether day–night cycles and light availability
influenced cheetah behavior, data on the time of the transition
between day and night and data on the phases of the moon cycle
were obtained from the United States Naval Observatory website
(http://aa.usno.navy.mil/data/). Discrimination between day and
night was based on the astronomical twilight, which is when the sun
is 18° below the horizon. Season was included as a variable because
cloud cover during the wet season is likely to influence moonlight
Figure 1
Definition of feeding point, feeding state, and feedings bout. F = feeding
and R = resting. Each compartment represents a 5-min interval.
Table 1
Number of labeled data points and the cross-validation of the combined SVM and HMM method that was used to translate the
activity data obtained from data-loggers, fitted to 6 free-ranging cheetahs, into 3 behavioral categories: feeding, mobile, and resting
Number of labeled data points
True positives
False positives
Number of predicted data points
Individuals
M1
M2
F1
F2
F3
F4
Feeding
Mobile
Resting
Total
Overall
Feeding
Mobile
Resting
Feeding
Mobile
Resting
Feeding
Mobile
Resting
Total
100
99
153
352
94%
93%
97%
93%
2%
9%
7%
4089
22 511
87 702
114 302
46
76
268
390
92%
52%
87%
100%
2%
24%
5%
4504
16 274
74 150
94 928
58
61
122
241
91%
95%
74%
98%
5%
0%
16%
2790
13 072
83 328
100 233
42
97
139
278
94%
100%
91%
95%
0%
7%
6%
2948
7306
68 179
78 434
31
81
171
283
84%
23%
75%
99%
6%
21%
16%
1862
9413
68 590
79 865
47
83
317
447
90%
66%
69%
99%
0%
2%
13%
2409
10 752
92 852
106 013
The true positives are the percentages with which the 3 different behaviors were detected correctly, whereas the false positives are the percentages with which
behaviors were falsely classified as one of the other behaviors.
Broekhuis et al. • Circalunar behavior of a diurnal carnivore
intensity at night and could therefore play a role in determining the
temporal distribution of cheetah behavior. There were therefore 3
predictor variables: diel phase (categorical with 2 levels: night and
day), season (categorical with 2 levels: wet and dry), and moonlight
intensity (a continuous variable rendering the proportion of the
moon that was visible where 0 = new moon and 1 = full moon).
As both lions and spotted hyaenas are predominately nocturnal
(Kruuk 1972; Schaller 1972; Hayward and Slotow 2009), the diel
phase (day vs. night) was used as a proxy for relative risk.
Statistical analysis
The proportion of time cheetahs spent feeding, mobile, or resting
in relation to the diel phase (day vs. night), season, and the moon
cycle were analyzed separately using Generalized Linear Models
(GLM) with a binomial error structure (Warton and Hui 2010).
The frequency of feeding bouts was similarly analyzed using GLMs
with a binomial error structure. The lengths of feeding bouts were
entered as a dependent variable in a linear model after being log
transformed to meet model assumptions. Individual cheetahs were
entered as a fixed blocking effect because of the small number of
individuals (Bates 2010). To analyze the proportion of time that
cheetahs spent in each behavioral state, 14 a priori candidate models were created for each of the behavioral states, including each
of the main effects (diel phase, season, and moonlight intensity)
and their interaction terms. Models were ranked using Akaike
Information Criterion corrected for small sample size. If one model
was clearly dominant (wi > 0.9), this was used, otherwise parameter
estimates were averaged across the models correcting for model
weights (wi; Burnham and Anderson 2002). Models and model
comparison statistics can be found in the Supplementary Appendix.
Statistical analyses were performed using the statistical software
R 2.14.2 (R Development Core Team 2012). Values are given as
mean ± 95% confidence interval.
1271
(Z = 0.152, P = 0.879; Figure 2). These results suggest that cheetahs actively shift their feeding behavior, and thus hunting behavior,
to the night during the dry season.
Moonlight intensity had no influence on the proportion of time
cheetahs spent feeding (Figure 3a) but did influence the proportion
of time for which cheetahs were mobile (Figure 3b). Higher levels
of moonlight were linked to a significant increase in mobile behavior at night (t124 = 4.724, P ≤ 0.001) and a significant decrease
in mobile behavior during the day (t124 = 3.308, P = 0.001). The
inverse was therefore true for the proportion of time cheetahs spent
resting: at night, cheetahs were more likely to rest when there was
less light present (t124 = −4.988, P ≤ 0.001) and more likely to be
resting during the day (t124 = 2.781, P = 0.006; Figure 3c).
Feeding frequency and length
Of the 910 feeding bouts, 12% took place at night, 14% in the
crepuscular transitions (i.e., dawn and dusk), and 73% took
place during the day. Feeding bouts at night lasted on average
62.29 ± 3.54 min and were significantly shorter than those during the day (97.15 ± 3.55 min; Z = 6.048, P ≤ 0.001; Figure 4).
Nocturnal feeding bouts were more frequent in the dry season than
in the wet season (Z = −3.235, P ≤ 0.001; Figure 5). During the dry
season, the night time frequency of feeding events increased significantly with increased moonlight (3-way interaction: Z = −2.275,
P = 0.023). However, moonlight intensity had no significant effect
on feeding length (Z = 1.277, P = 0.202). Season had a significant
effect on both the length and the frequency of feeding bouts (feeding length: Z = 1.972, P = 0.048; feeding frequency: Z = 11.349,
P ≤ 0.001; Figures 4 and 5). In the wet season, cheetahs fed more
frequently but feeding bouts where shorter (97.35 ± 2.52 min) compared with the dry season when cheetah fed less often but for longer (104.66 ± 4.34 min; Figure 4).
Discussion
Results
Temporal distribution of cheetah behavior
Cheetahs spent the majority of their time resting (84.5 ± 2.78%)
and only spent a small portion of their time feeding (3.0 ± 3.2%) or
mobile (12.5 ± 2.78%; Figure 2). Although cheetahs are predominantly diurnal, the results show that 45.2% of their overall mobile
and 32.5% of their overall feeding behavior occurred at night
(mobile: Z = 9.797, P ≤ 0.001; feeding: Z = 5.138, P ≤ 0.001).
Nonetheless, cheetahs still spent a greater portion of their time resting during the night than during the day (night: 94.2 ± 2.6%; day:
78.6 ± 2.3%; Z = 10.711, P ≤ 0.001; Figure 2).
The temporal distribution of cheetah behavior varied with
season. Cheetahs spent more time feeding during the wet season
compared with the dry season (Z = 2.424, P = 0.015), and there
was a significant interaction between the diel phase and season
(Z = 3.459, P ≤ 0.001), suggesting the day–night difference was not
the same in the wet and dry seasons. Cheetahs spent less time feeding during the day in the dry season compared with the wet season
(daydry: 3.4 ± 1.0%; daywet: 4.2 ± 0.5%; t124 = 2.238, P = 0.027) and
more time feeding at night in the dry season compared with the
wet season (nightdry: 1.2 ± 0.0%; nightwet: 0.4 ± 0.3; t124 = −3.384, P
≤ 0.001). Similarly, cheetahs were more mobile at night in the dry
season compared with the wet season (nightdry: 3.1 ± 2.4%; nightwet:
0.5 ± 0.2%; t124 = −2.770, P ≤ 0.001), whereas the amount of time
cheetahs were mobile during the day did not differ between seasons
Because animals are most at risk of costly encounters when feeding, not only because vigilance is reduced (Lima 1987), but also
because food resources attract kleptoparasites (Hunter et al. 2007a),
we hypothesized that cheetahs would not feed at night if nocturnal behavior was determined by predator avoidance. On the
other hand, we hypothesized that cheetahs would feed at night
if nocturnal behavior was driven by optimal hunting conditions.
We found that there were a high number of nocturnal feeding
events. Furthermore, an increase in nocturnal activity was offset by
decreased activity during the day, indicating an active shift toward
nocturnal behavior. Thus, our results indicate that the nocturnal behavior of predominantly diurnal cheetahs in our study can
be explained by optimal hunting conditions rather than predator
avoidance.
Cheetahs are visual hunters, thus we expected that the observed
nocturnal feeding behavior would be positively correlated with
moonlight intensity. Our results support this hypothesis, but only
during the dry season. This seasonal difference further illustrates
the importance of moonlight: cheetahs were more mobile and fed
more at night during the dry season compared with the wet season. During the wet season, the moon is frequently obscured by the
presence of clouds and thus the light conditions would be similar to
those during new moon in the dry season when cheetahs are barely
active at night. These results suggest that predators, especially those
that are visual hunters and use speed rather than ambush techniques, benefit from increased moonlight (Prugh and Golden 2014).
1272
Behavioral Ecology
Figure 2
Proportion of time cheetah in the Okavango Delta, Botswana (n = 6) spent feeding (black), mobile (dark gray), and resting (light gray) in relation to time of
day for the (a) dry and (b) wet season. The bar above the individual figures depicts the day–night length (day = gray, night = black) based on the astronomical
twilight on the solstices.
For example, the hunting efficiency of both short-eared owls (Asio
flammeus) and wolves (Canis lupus) was shown to increase with moonlight (Clarke 1983; Theuerkauf et al. 2003). Although these are
examples of nocturnal species, the same could be true for diurnal
cursorial predators. During moonlit nights, cheetahs would benefit
from increased visibility, whereas the darkness would allow them to
approach more closely to their prey. This may increase the rate of
successful hunts and/or decrease the chase distance and therefore
energy expenditure, similar to what has been found in wild dogs
(Lycaon pictus; Rasmussen et al. 2008). Furthermore, during moonlit
nights, impala (Aepyceros melampus), cheetahs’ primary prey (Purchase
and du Toit 2000; Hayward et al. 2006), tend to move into open
areas (Jarman and Jarman 1973). This may benefit cheetahs, which
are a high-speed, open habitat, cursorial predator.
Although hunting conditions may be optimal for cheetahs at
night when moonlight is present, this nocturnal behavior also suggests that cheetah behavior is not strictly shaped by the temporal
variation of risk. Rather, it is likely that avoidance of predators is
influenced by finer scale behavioral adaptations, resulting in partial,
rather than complete, temporal segregation (Carothers and Jaksić
1984; Kronfeld-Schor and Dayan 2003). For instance, direct cues,
such as calls of lions and spotted hyaenas, rather than indirect ones
such as day–night cycles may influence behavioral decisions (Kotler
et al. 1991; Durant 2000). Indeed, this was the conclusion of Kotler
et al. (1991) who showed that great Egyptian sand gerbils (Gerbillus
pyramidum) had a stronger response to the direct risk of owl predation
rather than the indirect risk associated with increased moonlight at
night. Similarly, cheetahs respond to the direct risk of encountering
other predators rather than the indirect risk (Broekhuis et al. 2013).
It is therefore plausible that, in general, the benefits of hunting at
night when lions and spotted hyaenas are active outweighs the risk
of negative encounters, but that hunting and feeding behavior ceases
when there is a direct risk of encountering predators (Durant 2000).
Accordingly, the present results indicate that feeding behavior itself
Broekhuis et al. • Circalunar behavior of a diurnal carnivore
1273
Figure 4
The length of cheetah feeding bouts at (a) different times of day and in (b)
different seasons.
Figure 3
Relationship between the moonlight intensity at night (0 = no moon, 1 = full
moon) and the proportion of time cheetah spent (a) feeding, (b) moving, and
(c) resting. Fitted lines are displayed ± 95% confidence intervals.
could be influenced by the increased likelihood of encountering nocturnal predators. Nocturnal feeding bouts were significantly shorter
than those during the day, implying either a higher rate of kleptoparasitism or that cheetahs left their kill sooner. Kleptoparasitism is energetically costly (Gorman et al. 1998) and would counterbalance the
possible energy gained by more efficient hunts at night. It is therefore possible that cheetahs either selected for smaller prey at night or
that they tried to spend less time on a kill when they were in a risky
situation. Indeed, Hunter et al. (2007b) found that cheetahs left kills
earlier when they were in an environment associated with more risk.
We also found that feeding length and frequency varied according to season with cheetahs feeding more frequently but for shorter
periods during the wet season compared with the dry season. This
Figure 5
Frequency of cheetah feeding bouts (adjusted for time) in relation to the
time of day and season.
may be explained by impalas calving at the beginning of the wet
season: during the wet season, young impalas make up 50% of the
total prey consumed by cheetahs (n = 38), whereas no young impalas were taken during the dry season (n = 20; Broekhuis F, unpublished data). Therefore, with the presence of smaller prey, cheetahs
may feed more frequently but for shorter periods of time.
There are other factors that could govern cheetahs’ nocturnal
behavior. First, temperature has been shown to influence cheetah activity (Cozzi et al. 2012). Because night time temperatures
are significantly lower than day time temperatures, it is possible
that cheetahs hunt at night to avoid high day time temperatures.
However, if this were the case in the current study, we should have
Behavioral Ecology
1274
found that night time feeding was more frequent in the hot, wet
season, whereas we found the opposite. Second, previous studies
have found that diurnal species shift their activity to the night as
a result of human activity during the day (Marnewick et al. 2006;
Rasmussen and Macdonald 2012). This is unlikely to be the case in
our study which was conducted in a wildlife area (Moremi Game
Reserve and the surrounding Wildlife Management Areas) where
anthropogenic pressures are minimal (Boggs 2000). We therefore
conclude that cheetahs naturally choose to be active at night rather
than being driven to do so by external factors. Although we realize that our sample size is small, these results may be characteristic
of cheetah behavior more broadly. Indeed, studies elsewhere have
noted similar nocturnal hunting activity by cheetahs (Stander 1990;
Bissett and Bernard 2007). It is plausible that this nocturnal behavior has not been quantified before because cheetahs are generally
recognized as being diurnal (Hilborn et al. 2012). Therefore, data
collection for cheetahs is typically biased toward daylight hours
(e.g., Durant 1998; Hunter et al. 2007b). As a result, the comparatively less frequent, nocturnal foraging and feeding behavior
may not have been documented in such detail because of lack of
sampling techniques. Traditional methods of behavioral data collection, especially direct observation, make it difficult to constantly
monitor animals, thus introducing a bias. Detailed studies on the
timing of behavior of elusive, far ranging species such as cheetahs that live at low densities or in remote areas that are difficult
to access have therefore been limited. Studies that have managed
to follow cheetahs continuously, including during the night, have
done so for comparatively short time intervals (14 d; Mills and
Biggs 1993; Hayward and Slotow 2009) and were therefore unable
to detect trends associated with long-term cycles such as the moon
and season. The development of data-loggers, that automatically
record and store animal activity data, has made it possible to investigate behavioral data, which was previously difficult or impossible
to obtain.
In summary, our results suggest that the nocturnal behavior of
cheetahs can be explained in terms of optimal hunting conditions.
In addition, cheetahs are more flexible and adaptive in their behavior than previously believed. The same could therefore be true for
other diurnal predators, such as African wild dogs. The results presented here illustrate that by taking longer term cycles (e.g., lunar
phase and season) into account, variations in animal behavior may
be detected. This study thereby sheds new light on cheetah behavior and contributes to our understanding of the importance of
moonlight and season on the behavior patterns of diurnal species.
Finally, our methodology and inclusion of lunar and seasonal variables, could be incorporated into studies on energetic budgets and
optimal diet models (e.g., Carbone et al. 1999), because our results
suggest that feeding events and frequencies vary over time.
Supplementary Material
Supplementary material can be found at http://www.beheco.
oxfordjournals.org/.
Funding
Tom Kaplan Prize Scholarship; Wilderness Wildlife Trust.
We thank the Botswana Ministry of Environment, Wildlife, and Tourism
for permission to conduct this research under permit number EWT 8/36/4
IV. We thank the Botswana Predator Conservation Trust for logistical support, A. Kusler for her help in the field, A. Wilson and S. Hailes for fruitful
discussions in the early stages of this research, and M. Valeix, P. Johnson,
N. Elliot, G. Cozzi, and 2 anonymous reviewers for commenting on previous drafts of the manuscript.
Handling editor: Shinichi Nakagawa
References
Bates DM. 2010. lme4: mixed-effects modeling with R. Springer [cited
2014 April 25]. Available from: http://lme4 r-forge r-project org/book.
Bishop CM. 2008. Pattern recognition and machine learning. New York:
Springer.
Bissett C, Bernard RTF. 2007. Habitat selection and feeding ecology of the
cheetah (Acinonyx jubatus) in thicket vegetation: is the cheetah a savanna
specialist? J Zool. 271:310–317.
Boggs LP. 2000. Community power, participation, conflict and development choice: community wildlife conservation in the Okavango region
of Northern Botswana. London: International Institute for Environment
and Development.
Broekhuis F, Cozzi G, Valeix M, McNutt JW, Macdonald DW. 2013. Risk
avoidance in sympatric large carnivores: reactive or predictive? J Anim
Ecol. 82:1098–1105.
Brown JS, Kotler BP. 2004. Hazardous duty pay and the foraging cost of
predation. Ecol Lett. 7:999–1014.
Burnham KP, Anderson DR. 2002. Model selection and multimodel
inference: a practical information-theoretic approach. New York:
Springer-Verlag.
Carbone C, Mace GM, Roberts SC, Macdonald DW. 1999. Energetic constraints on the diet of terrestrial carnivores. Nature. 402:286–288.
Caro TM. 1994. Cheetahs of the Serengeti: group living in an asocial species. Chicago: The University of Chicago Press.
Carothers JH, Jaksić FM. 1984. Time as a niche difference: the role of
interference competition. Oikos. 42:403–406.
Clarke JA. 1983. Moonlight’s influence on predator/prey interactions
between short-eared owls (Asio flammeus) and deermice (Peromyscus maniculatus). Behav Ecol Sociobiol. 13:205–209.
Cozzi G, Broekhuis F, McNutt JW, Turnbull LA, Macdonald DW,
Schmid B. 2012. Fear of the dark or dinner by moonlight? Reduced
temporal partitioning among Africa’s large carnivores. Ecology.
93:2590–2599.
Daan S, Aschoff J. 1975. Circadian rhythms of locomotor activity in captive
birds and mammals: their variations with season and latitude. Oecologia.
18:269–316.
Durant SM. 1998. Competition refuges and coexistence: an example from
Serengeti carnivores. J Anim Ecol. 67:370–386.
Durant SM. 2000. Living with the enemy: avoidance of hyenas and lions by
cheetahs in the Serengeti. Behav Ecol. 11:624–632.
Fenn MGP, Macdonald DW. 1995. Use of middens by red foxes: risk
reverses rhythms of rats. J Mammal. 76:130–136.
Gittleman JL. 1986. Carnivore brain size, behavioral ecology, and phylogeny. J Mammal. 67:23–36.
Gorman ML, Mills MG, Raath JP, Speakman JR. 1998. High hunting
costs make African wild dogs vulnerable to kleptoparasitism by hyaenas.
Nature. 391:479.
Griffin PC, Griffin SC, Waroquiers C, Mills LS. 2005. Mortality by moonlight: predation risk and the snowshoe hare. Behav Ecol. 16:938–944.
Grünewälder S, Broekhuis F, Macdonald DW, Wilson AM, McNutt JW,
Shawe-Taylor J, Hailes S. 2012. Movement activity based classification
of animal behaviour with an application to data from cheetah (Acinonyx
jubatus). PLoS ONE. 7:e49120.
Harrington LA, Harrington AL, Yamaguchi N, Thom MD, Ferreras P,
Windham TR, Macdonald DW. 2009. The impact of native competitors
on an alien invasive: temporal niche shifts to avoid interspecific aggression? Ecology. 90:1207–1216.
Hayward MW, Hofmeyr M, O’Brien J, Kerley GIH. 2006. Prey preferences
of the cheetah (Acinonyx jubatus) (Felidae: Carnivora): morphological limitations or the need to capture rapidly consumable prey before kleptoparasites arrive? J Zool. 270:615–627.
Hayward MW, Kerley GIH. 2008. Prey preferences and dietary overlap
amongst Africa’s large predators. S Afr J Wildl Res. 38:93–108.
Hayward MW, Slotow R. 2009. Temporal partitioning of activity in large
African carnivores: tests of multiple hypotheses. S Afr J Wildl Res.
39:109–125.
Broekhuis et al. • Circalunar behavior of a diurnal carnivore
Hilborn A, Pettorelli N, Orme CDL, Durant SM. 2012. Stalk and chase:
how hunt stages affect hunting success in Serengeti cheetah. Anim Behav.
84:701–706.
Hunter JS, Durant SM, Caro TM. 2007a. Patterns of scavenger arrival at
cheetah kills in Serengeti National Park Tanzania. Afr J Ecol. 45:275–281.
Hunter JS, Durant SM, Caro TM. 2007b. To flee or not to flee: predator
avoidance by cheetahs at kills. Behav Ecol Sociobiol. 61:1033–1042.
Jarman MV, Jarman PJ. 1973. Daily activity of impala. E Afr Wildl J.
11:75–92.
Klein DC, Coon SL, Roseboom PH, Weller J, Bernard M, Gastel JA, Zatz
M, Iuvone M, Rodriguez IR, Bégay V. 1997. The melatonin rhythm-generating enzyme: molecular regulation of serotonin N-acetyltransferase in
the pineal gland. Recent Progr Horm Res. 52:307–358.
Kock M, Meltzer D, Burroughs R. 2006. Chemical and physical restraint
of wild animals: a training and field manual for African species.
Johannesburg (South Africa): International Wildlife Veterinary Services.
Kotler BP, Brown JS, Hasson O. 1991. Factors affecting gerbil foraging
behavior and rates of owl predation. Ecology. 72:2249–2260.
Kotler BP, Brown J, Mukherjee S, Berger-Tal O, Bouskila A. 2010.
Moonlight avoidance in gerbils reveals a sophisticated interplay among
time allocation, vigilance and state-dependent foraging. Proc R Soc Lond
B Biol Sci. 277:1469–1474.
Kronfeld-Schor N, Dayan T. 2003. Partitioning of time as an ecological
resource. Ann Rev Ecol System. 34:153–181.
Kruuk H. 1972. The spotted hyena: a study of predation and social behaviour. Chicago: University of Chicago Press.
Laurenson MK. 1994. High juvenile mortality in cheetahs (Acinonyx jubatus)
and its consequences for maternal care. J Zool. 234:387–408.
Lima SL. 1987. Vigilance while feeding and its relation to the risk of predation. J Theor Biol. 124:303–316.
Lima SL, Bednekoff PA. 1999. Temporal variation in danger drives antipredator behavior: the predation risk allocation hypothesis. Am Nat.
153:649–659.
Marnewick KA, Bothma JdP, Verdoorn GH. 2006. Using camera-trapping
to investigate the use of a tree as a scent-marking post by cheetahs in the
Thabazimbi district. S Afr J Wildl Res. 36:139–145.
McNutt JW. 1996. Sex-biased dispersal in African wild dogs, Lycaon pictus.
Anim Behav. 52:1067–1077.
1275
Mendelson J, Vanderpost C, Ramberg L. 2010. Okavango Delta: floods of
life. Olympia Windhoek (Namibia): Raison cc.
Mills MGL, Biggs HC. 1993. Prey apportionment and related ecological
relationships between large carnivores in Kruger National Park. Zool
Sympos. 65:253–268.
Penteriani V, Kuparinen A, Mar Delgado M, Palomares F, López-Bao J,
Fedriani J, Calzada J, Moreno S, Villafuerte R, Campioni L, et al. 2013.
Responses of a top and a meso predator and their prey to moon phases.
Oecologia. 1–14.
Prugh LR, Golden CD. 2014. Does moonlight increase predation risk?
Meta-analysis reveals divergent responses of nocturnal mammals to lunar
cycles. J Anim Ecol. 83:504–514.
Purchase GK, du Toit JT. 2000. The use of space and prey by cheetahs in
Matusadona National Park, Zimbabwe. S Afr J Wildl Res. 30:139–144.
R Development Core Team. 2012. R: a language and environment for
statistical computing. Vienna (Austria): R Foundation for Statistical
Computing.
Rasmussen GSA, Gusset M, Courchamp F, Macdonald DW. 2008. Achilles’
Heel of Sociality Revealed by Energetic Poverty Trap in Cursorial
Hunters. Am Nat. 172:508–518.
Rasmussen GSA, Macdonald DW. 2012. Masking of the zeitgeber: African
wild dogs mitigate persecution by balancing time. J Zool. 286:232–242.
Schaller GB. 1972. The Serengeti lion: a study of predator-prey relations.
Chicago: University of Chicago Press.
Schoener T. 1974. Resource partitioning in ecological communities.
Science. 185:27–39.
Shawe-Taylor J, Cristianini N. 2004. Kernel methods for pattern analysis.
Cambridge: Cambridge University Press.
Sikes RS, Gannon WL. 2011. Guidelines of the American Society of
Mammalogists for the use of wild mammals in research. J Mammal.
92:235–253.
Stander PE. 1990. Notes on foraging habits of cheetah. S Afr Tydskr
Natuurnav. 20:130–132.
Theuerkauf J, Jędrzejewski W, Schmidt K, Okarma H, Ruczyński I,
Śnieżko S, Gula R. 2003. Daily patterns and duration of wolf activity in
the Białowieża Forest, Poland. J Mammal. 84:243–253.
Warton DI, Hui FKC. 2010. The arcsine is asinine: the analysis of proportions in ecology. Ecology. 92:3–10.