- Email: [email protected]
New insights into the huddling dynamics of emperor penguins
Animal Behaviour 110 (2015) 91e98
Contents lists available at ScienceDirect
Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav
New insights into the huddling dynamics of emperor penguins Ancel a, b, *, Caroline Gilbert c, Nicolas Poulin d, Michae €l Beaulieu e, Andre a, b Bernard Thierry Universit e de Strasbourg, Institut Pluridisciplinaire Hubert Curien, Strasbourg, France Centre National de la Recherche Scientifique, D epartement Ecologie, Physiologie et Ethologie, Strasbourg, France c Universit e Paris-Est, Ecole Nationale V et erinaire d'Alfort, Centre National de la Recherche Scientifique, Mus eum National d'Histoire Naturelle, Paris, France d CeStatS, Institut de Recherche Math ematique Avanc ee, Universit e de Strasbourg, Centre National de la Recherche Scientifique, Strasbourg, France e Zoological Institute and Museum, University of Greifswald, Greifswald, Germany a
b
a r t i c l e i n f o Article history: Received 13 March 2015 Initial acceptance 9 April 2015 Final acceptance 31 August 2015 Available online MS. number: 15-00215R Keywords: aggregation Antarctica Aptenodytes forsteri energy saving environment huddling social thermoregulation
Social thermoregulation is a cooperative strategy in which animals actively aggregate to benefit from the warmth of conspecifics in response to low ambient temperatures. Emperor penguins, Aptenodytes forsteri, use this behaviour to ensure their survival and reproduction during the Antarctic winter. An emperor penguin colony consists of a dynamic mosaic of compact zones, the so-called huddles, included in a looser network of individuals. To maximize energy savings, birds should adjust their huddling behaviour according to environmental conditions. Here, we examined the dynamics of emperor penguin aggregations, based on photo and video records, in relation to climatic factors. Environmental temperature, wind and solar radiation were the main factors contributing to huddle formation. The analysis of individual movements showed that birds originating from loose aggregations continually joined huddles. Sometimes, a small number of birds induced a movement that propagated to the entire huddle, causing its breakup within 2 min and releasing birds, which then integrated into looser aggregations. Different parts of the colony therefore appeared to continually exchange individuals in response to environmental conditions. A likely explanation is that individuals in need of warmth join huddles, whereas individuals seeking to dissipate heat break huddles apart. The regular growth and decay of huddles operates as pulses through which birds gain, conserve or lose heat. Originally proposed to account for reducing energy expenditure, the concept of social thermoregulation appears to cover a highly dynamic phenomenon that fulfils a genuine regulatory function in emperor penguins. © 2015 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Coloniality is the aggregation of individuals of the same species on a breeding ground remote from foraging sites (Kharitonov & Siegel-Causey, 1988). This kind of social reproduction is used by many species of vertebrates, and especially by seabirds among which more than 95% nest in colonies (Brown, Stutchbury, & Walsh, 1990). Living within a colony may entail costs in terms of intraspecific competition, spread of pathogens or travel distances (e.g. Kharitonov & Siegel-Causey, 1988; Krause & Ruxton, 2002). To be adaptive, coloniality has to bring fitness benefits that offset these zilly, & Pagel, 1988; costs (e.g. Danchin & Wagner, 1997; Dubois, Ce Siegel-Causey & Kharitonov, 1991). For some species, one main advantage of group living is social thermoregulation (Gilbert et al., 2010). This is a cooperative
* Correspondence: A. Ancel, CNRS/IPHC/DEPE, 23 rue Becquerel, 67087 Strasbourg, France. E-mail address: [email protected] (A. Ancel).
strategy that operates through both physiological and behavioural processes. It can be defined as the active aggregation of individuals to benefit from the warmth of conspecifics in response to low temperatures (Alberts, 1978; Gilbert et al., 2010; Martin, Fiorentini, & Connors, 1980). Huddling animals maximize energy savings by reducing the body surface area exposed to cold, thus decreasing heat loss, and by warming their surrounding environment. Huddling occurs in vertebrates as diverse as rodents, bats, primates and penguins (e.g. Gilbert, Blanc, Le Maho, & Ancel, 2008; Hayes, 2000; Ostner, 2002; Willis & Brigham, 2007). Huddles of emperor penches of king penguin chicks, guins, Aptenodytes forsteri, and cre Aptenodytes patagonicus, can involve thousands of birds, which cooperate to share warmth and decrease predation risk (Ancel, Visser, Handrich, & Le Maho, 1997; Gilbert, Robertson, Le Maho, Naito, & Ancel, 2006; Le Bohec, Gauthier-Clerc, & Le Maho, 2005; vost, 1961). Pre Huddling is probably a key factor allowing the colonization of particularly cold habitats by endotherms. Emperor penguins
http://dx.doi.org/10.1016/j.anbehav.2015.09.019 0003-3472/© 2015 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
92
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
represent the most extreme example, as they are the only animals able to breed during the Antarctic winter, far from the open sea vost, 1961). Their thermoneutral zone ranges where they feed (Pre from 10 to þ20 C, and they can withstand negative ambient temperatures only to a certain extent before increasing their metabolism to generate heat (Le Maho, Delclitte, & Chatonnet, 1976; Pinshow, Fedak, Battles, & Schmidt-Nielsen, 1976). However, when conditions deteriorate further (at 25 C, the probability of huddling is ca. 80% for a wind speed of 0 m/s and 98% for a wind speed of 15 m/s, Gilbert, Robertson, Le Maho, & Ancel, 2008), behavioural strategies complement physiological strategies, which allows emperor penguins to save energy (Le Maho, 1977). Under similar weather conditions, free-ranging males that huddled maintained a metabolic rate 16% lower than captive males that were experimentally prevented from forming dense aggregations (Ancel et al., 1997). When wind chill increases the cooling rate, emperor penguin vost, colonies move downwind (Gilbert, Robertson et al., 2008; Pre 1961; Robertson, 1990). By contrast, when wind chill decreases, colonies move against the wind, presumably because birds tend to return to their usual sites to avoid areas of unstable ice. Therefore, colonies move back and forth under the influence of winds (Gilbert, Robertson et al., 2008). It is known that a major part of energy saving in loosely grouped birds is from mutual wind protection (Ancel et al., 1997). Low ambient temperatures, however, appear more influential than wind speed in inducing the formation of huddles (Gilbert, Le Maho, Perret, & Ancel, 2007; Gilbert, Robertson et al., 2008), but we do not know to what extent other meteorological factors can affect the grouping patterns of emperor penguins. A colony of emperor penguins is spatially heterogeneous, consisting of several parts of varying densities (Gilbert et al., 2006). In the tighter aggregations called huddles, density can reach vost, 1961), individuals step in a synchronized 8e10 birds/m2 (Pre way (Gerum et al., 2013; Zitterbart, Wienecke, Butler, & Fabry, 2011) and the birds most exposed to the wind move along the opposite vost, 1961; flank of the huddle for protection (Birr, 1968; Pre Robertson, 1990; Waters, Blanchette, & Kim, 2012). As heat loss is limited within huddles (as heat can dissipate only by inhalation of cold air and through the head) ambient temperature can reach 37.5 C (Gilbert et al., 2006), well above the birds' þ20 C upper critical temperature (Le Maho et al., 1976; Pinshow et al., 1976). As a consequence, birds face the paradox that in a cold physical environment they sometimes need to dissipate excess heat (Gilbert et al., 2007, 2006). At present, we do not know how emperor penguins deal with these apparently contradictory requirements since the factors influencing the process of huddle formation and breakup have never been studied in detail with regard to environmental variables. Most studies have examined the effects of weather conditions on emperor penguin colony structure only during incubation, but since energetic constraints are likely to differ between the breeding stages it is important to examine them also during pairing and chick rearing. For instance, penguins have to fast for a prolonged time during incubation when weather conditions are most severe, while they only stay in the colony for short periods when rearing chicks, when weather conditions are more favourable. Consequently, forming huddles may be more critical during incubation relative to other breeding stages to maximize energy savings. To understand the dynamics of social thermoregulation in emperor penguins according to environmental variables (ambient temperature, wind speed, wind direction, relative humidity, atmospheric pressure or solar radiation), we investigated the huddling behaviour of an entire colony during pair formation, incubation and chick rearing, that is, from winter to summer. We
aimed to test the following predictions: (1) as dense aggregations protect emperor penguins from inclement weather conditions and help individuals to save energy, meteorological variables that have the potential to lower body temperature should result in an increase in huddle formation and duration, (2) as social thermoregulation is based upon the need of individuals to save energy by forming huddles and the need to shed excess heat by leaving huddles, this should lead to the regular formation and breakup of huddles, and (3) as a colony of emperor penguins is made of aggregations of different densities, individuals should move between loose aggregations and huddles to control body temperature. METHODS Subjects ologie We studied the emperor penguin colony of Pointe Ge lie, Antarctica, 66 400 S, Archipelago (Dumont d'Urville, Terre Ade 140 010 E) where about 3000 pairs breed each year on the fast-ice vost, 1961). We between islands and the Antarctic continent (Pre collected data in 2005, 2006 and 2008, over three reproductive periods: pairing (from arrival of the birds in late April to mid-May), incubation (from mid-May to mid-July) and chick rearing (from mid-July to mid-September). Fieldworkers counted the birds in the colony at each season, once a week depending on the weather conditions (good visibility, low wind). There were a maximum of 3253, 3350 and 3160 breeding males during the incubation periods in 2005, 2006 and 2008, respectively. Definitions To analyse the distribution of emperor penguins within the colony, we differentiated between two different grouping patterns: huddles and loose aggregations. A huddle was defined as a part of a group in which individuals were closely assembled; typically, individuals had flippers held against the body and the head tucked into the shoulders, and each bird hid its beak between the necks of others. Note that movements within a huddle are limited, birds being only able to move the head or to make limited steps in a synchronized way (Gerum et al., 2013; Zitterbart et al., 2011). Sometimes a huddle broke down when its members became more active (Fig. 1). A group could contain one or several huddles. Loose aggregations consisted of all the individuals in a group that did not belong to huddles. These birds were stationary or mobile, and had limited if any physical contact with their neighbours or were apart from each other. They engaged in different activities such as grooming, lifting the abdominal skin fold covering the egg, consuming snow or performing courtship displays when females were present. During the pairing period, both males and females were present in the huddles but during the incubation period, only males were present. Climatic and Seasonal Influences on Huddling Patterns To investigate huddling patterns, we photographed the colony every day (weather permitting) from late April to mid-September 2008. Several photos were taken within three different time windows (morning: 0930e1130 hours; midday: 1130e1330 hours; afternoon: 1330e1530 hours; local time i.e. UTC þ 10 h), each time from three vantage points located on nearby hills (from two points on Le Mauguen Island and from one point on Rostand Island). The different photographs of the colony were collated to get a single picture of the entire colony using PTGui software (www.ptgui.com). Each picture was visually analysed on a computer screen using the zoom function. This led to a database of 118 panoramic pictures: 20
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
93
Figure 1. Grouping patterns in emperor penguins during the incubation period: (a) huddles on the centre and bottom of the picture, with loose aggregations around them; (b) the huddle in the centre starts to break on the right side; (c) after breakup birds reach a state of loose aggregation.
during pairing (morning: 11; midday: 0; afternoon: 9), 60 during incubation (21, 21, 18) and 38 during chick rearing (14, 11, 13). We obtained meteorological data from the permanent weather te o France at Dumont d'Urville located on Petrel station of Me
Island, at 43 m in height and 500 m from the colony of emperor penguins. Because temperatures and wind speed have already been described as being lower at the colony site than at the meteorological station, they were corrected for the purpose of our study
94
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
according to the equations given by Gilbert et al. (2007). We calculated a daily mean from measures taken every 3 h: air temperature, atmospheric pressure, relative humidity at the corresponding air temperature, wind speed and wind direction; solar radiation was summed on a daily basis (Appendix Table A1). Duration of Huddles, Breakups and Movements of Individuals
huddles and mean number of individuals per huddle. Predictor variables were meteorological factors (see above), reproductive and daytime periods, and their interactions. We considered only simple interactions since the models did not converge with multiple interactions. Each quantitative predictor variable was scaled before being used in the models. As the number of huddles corresponded to count data, generalized linear models (GLM) using a quasiPoisson distribution were fitted. When using the quasi-Poisson family, the Akaike information criterion is not available; hence we used likelihood ratio tests (LRT) to select the best model (Lewis, Butler, & Gilbert, 2011). To obtain P values for each variable, only interactions were removed according to LRT. We applied this procedure for all models. With regard to the mean number of individuals per huddle, residuals followed a normal law when data were log-transformed, so we used a linear model on the logtransformed mean number of individuals per huddle. For the logtransformed mean number of individuals per huddle, the best model was selected using LRT. Post hoc comparisons were made using the R package ‘multcomp’ with Tukey corrections (Hothorn, Bretz, & Westfall, 2008). We analysed the movements of individuals between huddles and loose aggregations using the Wilcoxon signed-ranks test, exact procedure. The significance level was set at 0.05. Average values are given as means and standard errors of the mean (SEM).
To investigate the dynamics of huddles and the movements of individuals during daylight, we videotaped the colony during the incubation period using a digital camera between 1000 and 1600 hours local time, i.e. from sunrise to sunset. We collected 22 h of recording over 7 days in 2005 (from 6 to 12 July) and 46 h over 16 days in 2006 (between 25 May and 28 June). Images were visually analysed frame by frame. A bird was considered as moving when it advanced at least three steps. It could walk, stop or join a huddle. It could also leave a huddle and reach a loose aggregation. We estimated the duration of huddles from six video clips, recorded in June 2006, which displayed the entire colony for at least 3 h. We counted the huddles at 15 min intervals (N ¼ 13) using scan sampling. We also counted the breakups occurring across the whole duration of videos. The mean duration of huddles was defined as the ratio of the total number of huddles over all scans in the colony divided by the number of breakups, thus providing measures given in number of 15 min intervals. We assessed the time individuals spent in huddles from data recorded with data loggers (Mk9, Wildlife Computers, Redmond, WA, U.S.A.) deployed on 13 males in 2005 and 2006. The tags, which were glued to their lower back, recorded ambient temperature and light intensity every 5 s in 2005 and every 2 s in 2006 (Ancel, Beaulieu, Le Maho, & Gilbert, 2009). An emperor penguin was deemed to be in a huddle when its back was completely covered by another bird behind it and the recorded light intensity dropped below 40 arbitrary units (a.u.); light exceeded 120 a.u. during the day and was 60e80 a.u. at night. When a bird entered a huddle the temperature increased at the same time as the light level dropped and vice versa when it left the huddle (see Ancel et al., 2009). We examined the origin of breakups from 20 huddles for which the start of breakup was videotaped. We assessed the temporal dynamics of breakups from nine video clips in which the whole process was recorded. We then selected six breakups for which we could follow individuals during the 5 min following breaking up in order to investigate the outcome of breakups. To assess the temporal dynamics of breakups, we recorded for every bird belonging to the huddle whether they were motionless or showing any sign of activity at 20 s intervals. We assessed the expansion of a huddle during a breakup by measuring the maximal length and width of the area occupied by the birds before and after breakup. To estimate the area covered by a huddle we assumed that it was shaped like an ellipse. In a final step, we determined the time elapsed between a breakup and the subsequent re-establishment of a huddle (i.e. at least 20 individuals) in the same location by monitoring the outcomes of 23 breakups for as long as allowed by video records. To study the movements of individuals between huddles and loose aggregations, we randomly selected 11 sequences from five video clips that displayed both a loose aggregation and more than one huddle, with no breakup occurring during these sequences. We counted the birds in huddles and loose aggregations, and the number of these moving from one type of grouping to another.
The number of huddles increased when air temperature and wind speed diminished and when solar radiation rose and also changed according to wind direction (Table 1). We additionally found a negative interaction between wind direction and relative humidity (Table 1). The mean number of individuals per huddle increased when air temperature or solar radiation decreased, and when wind speed increased (Table 1). The reproductive period had a significant effect on the two target variables (Table 1). Significant interactions appeared between reproductive season and climatic factors: with air temperature for number of huddles and with solar radiation for mean number of individuals per huddle (Table 1). Additionally, an interaction appeared between daytime period and wind direction for the number of huddles (Table 1). Huddling patterns by periods are presented in Table 2. The two target variables varied as a function of reproductive period. There were fewer huddles during the incubation period (3.6 ± 1.2) than during the pairing (6.5 ± 1.3) and chick-rearing periods (11.8 ± 2.1; Tukey tests: pairing versus incubation: P < 0.001; incubation versus chick rearing: P < 0.001; pairing versus chick rearing: P ¼ 0.144). The mean number of individuals per huddle was higher during the incubation period (897 ± 35) than the chick-rearing period (123 ± 24; Tukey test: P ¼ 0.013), but no significant differences appeared between pairing (246 ± 87) and incubation (P ¼ 0.256) or chick-rearing periods (P ¼ 0.630).
Statistical Analyses
Duration and Breakup of Huddles
We analysed daily data on climatic and seasonal influences with R 2.15.0 (R Core Team, 2012). Target variables were number of
Birds spent on average 62.1 ± 2.9 min (N ¼ 50 huddles) and 41.0 ± 0.6 min (N ¼ 185 huddles) in huddles in 2005 (7 days) and
Ethical Note The study was approved by the ethics committee of the Institut Polaire Français Paul-Emile Victor (IPEV) and the Terres Australes et Antarctiques Françaises (TAAF). RESULTS Climatic and Seasonal Influences on Huddling Patterns
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
95
Table 1 Results of GLM analyses testing the influence of climatic factors, reproductive and daytime periods and their interactions on huddling patterns Predictor variable
Number of huddles
Air temperature Atmospheric pressure Relative humidity Solar radiation Wind speed Wind direction Reproductive period Daytime period Air temperature ) reproductive period Wind direction ) relative humidity Wind direction ) daytime period Solar radiation ) reproductive period
Mean number of individuals per huddle
Adjusted generalized VIF
Estimate
Decision statistics (t or F)
P
Adjusted generalized VIF
Estimate
Decision statistics (t or F)
P
2.13 1.41 1.37 1.43 1.39 2.44 1.82 1.07 1.82 1.76 3.90
0.348 0.118 0.006 0.141 0.172 0.438
2.80 1.74 0.08 2.02 2.10 3.94 30.30 0.99 19.4 3.75 15.4
0.006 0.085 0.935 0.046 0.039 <0.001 <0.001 0.374 <0.001 <0.001 <0.001
1.76 1.38 1.30 2.75 1.40 1.32 7.63 1.05
0.472 0.115 0.036 0.939 0.416 0.188
3.38 1.07 0.36 4.34 3.86 1.80 24.2 0.894
0.001 0.288 0.721 <0.001 <0.001 0.074 <0.001 0.412
14.30
<0.001
0.305
2.20
VIF ¼ variance inflation factor. Estimates and t values are provided for the effects of climatic factors and F values for the effects of periods. Significant results are indicated in bold.
Table 2 Huddling patterns according to reproductive and daytime periods (mean ± SEM) Huddling patterns
Pairing
Number of huddles Number of individuals per huddle
Incubation
Chick rearing
Morning (11)
Midday (0)
Afternoon (9)
Morning (21)
Midday (21)
Afternoon (18)
Morning (14)
Midday (11)
Afternoon (13)
6.2±1.7 262±87
e e
6.8±0.9 226±88
4.3±1.4 798±39
3.5±1.3 1106±38
3.0±0.9 769±26
12.7±2.1 186±16
13.5±2.6 103±27
9.4±1.6 73±29
Number of days is given in parentheses.
Percentage of moving individuals
100
80
60
40
20
0 –160 –120
–80
–40 0 40 80 20 s time intervals
120
160
200
Figure 2. Dynamics of the breakup process: mean percentage of moving individuals over time (at time interval 0 the number of moving individuals passes from 50 to over 50).
number of 599 ± 243 birds). Most dispersing birds joined loose aggregations (99.7 ± 0.2% of individuals; Fig. 1c) rather than other huddles (0.3 ± 0.2%). Huddles did not form again immediately after breakup; they re-established at the same location within 1 min in only 12% of cases, while after 40 min 80% of huddles had reformed (Fig. 3).
Movements of Individuals Between Huddles and Loose Aggregations The percentage of individuals transferring from a loose aggregation to a huddle (2.1 ± 0.6% of individuals staying in loose aggregations per min) was significantly higher than the percentage of individuals leaving a huddle to join a loose aggregation (0.3 ± 0.1%
Cumulative percentage of reformed huddles
2006 (16 days), respectively. Huddles could last less than one, and more than 13, 15 min intervals (i.e. more than ca. 3 h). We found a mean duration of 6.1 ± 1.9 min at 15 min intervals per huddle, i.e. about 90 min. Over the 20 huddles for which the start of the breakup was videotaped, 19 started near the edge of the huddle (see Fig. 1b) and one from the centre. One breakup was induced by a conflict between two individuals, six by the departure of several individuals, and no visible cause was detected for the remaining 13 events. During breakup the percentage of moving individuals rose from 2.7 ± 0.7% to 99.9 ± 0.1% within 2 min for a mean number of 434 ± 88 birds (Fig. 2). With regard to spatial dynamics, the ratio between the area occupied by individuals in a huddle and that occupied by these individuals after breakup was 1.8 ± 0.1 (mean
100
80
60
40
20
0
1
5
10 20 30 40 60 80 100 120 140 160 Time (min)
Figure 3. Latency between breakup and re-establishment of huddling in the same location of the colony (N ¼ 23 huddles).
96
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
of individuals staying in loose aggregations per min; Wilcoxon test: N ¼ 11 sequences, z ¼ 2.76, P ¼ 0.003). Focusing on huddles, we then determined from where individuals came when aggregating to huddle, or where they went when they left a huddle. We compared the flow of individuals entering and leaving a huddle within 1 min. The percentage of birds entering the huddle was three times as large as the percentage of individuals leaving it (75.0 ± 4.7% and 25.0 ± 4.7% per min, respectively; Wilcoxon test: N ¼ 22 huddles, z ¼ 3.47, P < 0.001). Among those integrating into a huddle, the percentage of birds coming from a loose aggregation was 10 times as large as the percentage of birds coming from a huddle (90.9 ± 8.3% and 9.1 ± 8.3% per min, respectively; Wilcoxon test: N ¼ 8 huddles, z ¼ 2.52, P ¼ 0.008). Among those leaving a huddle, the percentage of birds entering a loose aggregation was four times as large as the percentage of birds joining a huddle (80.0 ± 4.8% and 20.0 ± 4.8% per min, respectively; Wilcoxon test: N ¼ 8 huddles, z ¼ 2.52, P ¼ 0.008). To test whether the transfer of penguins was influenced by wind direction, we compared for 21 huddles the percentages of birds aggregating on the side of the huddle exposed to the wind versus the leeward side (62.4 ± 5.3% and 37.6 ± 5.3%, respectively). This did not produce a statistically significant difference (Wilcoxon test: N ¼ 21, z ¼ 1.40, P ¼ 0.168). No significant differences emerged either when comparing the percentages of birds leaving the exposed side versus the leeward side (36.6 ± 8.2% and 63.4 ± 8.2%, respectively; Wilcoxon test: N ¼ 12, z ¼ 1.36, P ¼ 0.203). DISCUSSION Overall, our results show that air temperature, wind and solar radiation were the main drivers pushing emperor penguins to gather in huddles. The influence of weather conditions on huddle dynamics was reflected by variable huddling patterns across breeding stages. Moreover, individuals regularly shifted between aggregations of different densities; they slowly moved from loose aggregations to huddles, while they rapidly left huddles following breakups. Climatic and Seasonal Influences on Huddling Patterns The number of huddles and the number of individuals per huddle increased when air temperature decreased. Moreover, when wind speed rose or solar radiation diminished, the mean number of individuals per huddle increased, while the number of huddles decreased, meaning that huddles got larger. Other climatic factors such as wind direction and relative humidity also favoured the formation of huddles. The effects of air temperature on the number of huddles is surprising, as one would expect a few large huddles to enhance heat conservation when the temperature sinks. One reason may be that when temperatures are low, penguins form huddles with the closest individuals (they do not look for a large pre-existing huddle), so that a large number of small huddles form. In contrast, when conditions are milder, penguins may take time to find a pre-existing huddle. The breeding stages corresponded to different physiological and demographic conditions. In the pairing period, females were still present and the colony remained dispersed. During incubation, males had to minimize heat loss to survive fasting while incubating vost, 1961). During chick rearing, chicks (Gilbert et al., 2006; Pre were protected against cold and predation, and were fed alternately by the parents. Moreover, the pairing, incubation and chick-rearing periods corresponded to different climatic seasons (Pierard & Pettre, 1993). In our study, the number of huddles decreased during incubation relative to the pairing and the chick-rearing periods.
In parallel, the largest number of penguins per huddle was found during incubation (up to 1100 individuals) while during chick rearing the number was lowest (fewer than 200 birds). As a corollary, loose aggregations were more frequent during chick rearing than at other periods. It may be hypothesized that this seasonal huddling pattern reflects the number of individuals present in the colony. Indeed, during the pairing period, males and females are present in the colony while only males are present during incubation, and adults are sporadically present at the end of the chickrearing period. The number of huddles would therefore be expected to be about twice as large during the pairing period as during incubation, and lower during incubation than during the chick-rearing period. Although this expected difference was observed between the pairing period and incubation, it was not found between the chick-rearing period and other periods. This indicates that factors other than the number of birds influence huddling patterns in emperor penguins. Accordingly, we still found significant effects of air temperature, solar radiation, wind speed and wind direction on huddling patterns after controlling for reproductive season. Duration and Breakup of Huddles The average time spent by individuals in a huddle fluctuated around 50 min, and huddles themselves lasted between a dozen of minutes to several hours. These numbers compare well with a huddle duration of 1.6 ± 1.7 h (mean ± SD), as previously reported in the same colony (Gilbert et al., 2006). These results indicate that emperor penguins packed in huddles remain in this state only for relatively short periods. By following the fate of huddles, we were able to highlight their breaking-up processes. Breakups could originate from the departure of some individuals or from agonistic behaviours between neighbours, but in most cases no visible cause was found (see also Mougin, 1966; Robertson, 1990). In accordance with our second prediction, breakups appeared as a regular process through which emperor penguins collectively left huddles. Breakups generally started close to the edge of the huddle, then propagated within 2 min to the entire huddle. Similarly, waves moving across the huddle appear to be initiated by a few individuals (Gerum et al., 2013). According to the thermoregulatory hypothesis, breakups were expected to occur more frequently in the middle of huddles, where temperature may be maximal. However, our results indicate that this may not be the case, which suggests that ambient temperatures may not necessarily increase linearly from the edge to the centre of the huddle. Another explanation is that penguins in the centre are stuck in the huddle, while penguins at the periphery can move enough to initiate huddle breakups more frequently. Breakups expanded until the huddles reached an area almost twice as large as that they initially occupied. Within 5 min after breakup, the previous members of a huddle seldom joined another huddle, meaning that almost all of them then belonged to the loosely aggregated part of the colony. Breakups therefore appear as the very process by which emperor penguins can release themselves from mutual body contact and dissipate excess heat. The breakup of huddles is sometimes accompanied by a haze of warm air rising over the colony, which indicates a significant dissipation of heat (Robertson, 1990). A previous study showed that penguins leave huddles when their body temperature reaches 37.5 C (Ancel et al., 2009). Those penguins whose skin temperature (37.5 C) is close to the animal's deep body temperature (Gilbert et al., 2006) could aim to find a lower external temperature to cool down. Related to this, we also observed that some birds leaving a huddle consumed fresh snow. Finally, other motives unrelated to thermoregulation may lead birds to initiate a breaking-up process, such
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
as the need to move freely or avoid agonistic behaviours within the huddle. Another important factor that may lead penguins to leave the huddle is that incubating birds need to regularly turn their egg to ensure the proper development of the embryo and thereby hatching success (Elibol & Brake, 2004). Movements of Individuals Between Huddles and Loose Aggregations The analysis of individual movements showed that most movements were directed from loose aggregations to huddles. This indicates that huddles were growing when observations were conducted. Individuals continually percolate from loose aggregations to huddles, while they join loose aggregations through the abrupt breaking up of huddles. To our knowledge, no equivalent process has ever been reported in another species, but it might also che period (Le Bohec occur in king penguin chicks during the cre et al., 2005). It is worth noting that some emperor penguins located on the outer edge of huddles can leave them without disturbing their neighbours. These birds also preferred to integrate into loose aggregations rather than join other huddles, which is consistent with the thermoregulatory hypothesis. If animals look for colder conditions, it is understandable that they do not immediately join another huddle but need time outside huddles to dissipate heat. It might happen that birds located on the outer edge of huddles need to reduce their exposure to the cold. Some authors have proposed that birds that are most exposed to the wind regularly break away to reach the sheltered side of the huddle (Birr, 1968; vost, 1961; Robertson, 1990). However, we did not find eviPre dence for such a phenomenon in our study. The birds did not leave the leeward side more often than the exposed side, and they did not show any significant preference for one side or the other when joining a huddle. The number of departures from huddles was negligible compared to the number of arrivals. When a bird integrates into a huddle, it is quickly surrounded by other individuals aggregating behind it, suggesting that it probably does not need to choose a particular side. Conclusion and Perspectives In agreement with our third expectation, our results support the view that emperor penguin colonies represent dynamic networks, the different parts of which continually exchange individuals in response to environmental conditions (Gilbert et al., 2006). It appears that individuals in need of warmth join huddles, and loose aggregations continually feed huddles. Conversely, huddles feed loose aggregations back, albeit in a discontinuous way. This temporal difference between joining and leaving a huddle presumably comes from the fact that individuals in loose aggregations differ in their need for warmth, while individuals in huddles simultaneously need to dissipate heat. The regular growth and decay of huddles therefore appear as pulses through which birds conserve or lose heat. Originally proposed to account for the reduction in energy expenditure made possible by social grouping, the concept of social thermoregulation appears to cover a dynamic phenomenon that fulfils a genuine regulatory function in emperor penguins. We acknowledge that our results hold for one colony of emperor lie, and caution is needed when generalizing penguins in Terre Ade to other colonies. It would be necessary to verify that our results are not due to particular meteorological conditions, and that similar processes occur in other sites too. For instance, the size of the ologie compared to other colony is relatively small at Pointe Ge locations. Some colonies are even smaller, however, making social thermoregulation more difficult. How the size of the colony can affect huddling patterns in emperor penguins remains to be
97
examined. Moreover, it should be kept in mind that our visual records necessarily took place during periods of good visibility, that is, under relatively good weather conditions, and when the safety rules of the French station permitted fieldwork. Because of low visibility, we could not make records in the most critical climatic conditions faced by the birds, and in particular at night or during severe blizzards for evident safety reasons. The next step will be to use covered infrared cameras to continuously monitor huddle formation and breakup during blizzards and at night, by time-lapse recording. This may be complemented by the deployment of operated stations such as wireless networks or rovers (Le Maho et al., 2014), able to identify an animal previously equipped with a transponder and to record its position accurately by using a GPS. Finally, developing software able to automatically follow the same individuals for long periods (Gerum et al., 2013; McCafferty et al., 2013; Zitterbart et al., 2011) offers the opportunity to improve analyses of huddling dynamics in colonies of emperor penguins. Acknowledgments Fieldwork was financially and logistically supported by the Institut Polaire Français Paul-Emile Victor (IPEV) and the Terres te o France generously Australes et Antarctiques Françaises. Me provided meteorological data. We thank Antoine Dervaux and le ne Petit and Vincent Anne-Mathilde Thierry for fieldwork, and He Martin for data analysis. Finally, we particularly thank the two anonymous referees who provided valuable comments and contributed to a significant improvement of this paper. References Alberts, J. R. (1978). Huddling by rat pups: group behavioral mechanisms of temperature regulation and energy conservation. Journal of Comparative and Physiological Psychology, 92, 231e245. Ancel, A., Beaulieu, M., Le Maho, Y., & Gilbert, C. (2009). Emperor penguins mates: keeping together in the crowd. Proceedings of the Royal Society B: Biological Sciences, 276, 2163e2169. Ancel, A., Visser, H., Handrich, Y., & Le Maho, Y. (1997). Energy saving in huddling penguins. Nature, 385, 304e305. tho-e cologiques a la colonie de manchots empereurs Birr, G. (1968). Observations e ologie en 1966. L'Oiseau et la Revue Française d'Ornithologie, 38, de Pointe Ge 53e88. Brown, C. R., Stutchbury, B. J., & Walsh, P. D. (1990). Choice of colony size in birds. Trends in Ecology & Evolution, 5, 398e403. Danchin, E., & Wagner, R. H. (1997). The evolution of coloniality: the emergence of new perspectives. Trends in Ecology and Evolution, 12, 342e347. zilly, F., & Pagel, M. (1988). Mate fidelity and coloniality in waterbirds: Dubois, F., Ce a comparative analysis. Oecologia, 116, 433e440. Elibol, O., & Brake, J. (2004). Identification of critical periods for turning broiler hatching eggs during incubation. British Poultry Science, 45, 631e637. Gerum, R. C., Fabry, B., Metzner, C., Beaulieu, M., Ancel, A., & Zitterbart, D. P. (2013). The origin of traveling waves in an emperor penguin huddle. New Journal of Physics, 15. njp483133. Gilbert, C., Blanc, S., Le Maho, Y., & Ancel, A. (2008). Energy saving processes in huddling emperor penguins: from experiments to theory. Journal of Experimental Biology, 211, 1e8. Gilbert, C., Le Maho, Y., Perret, M., & Ancel, A. (2007). Body temperature changes induced by huddling in breeding male emperor penguins. American Journal of Physiology, 292, R176eR185. Gilbert, C., McCafferty, D., Le Maho, Y., Martrette, J. M., Giroud, S., Blanc, S., et al. (2010). One for all and all for one: the energetic benefits of huddling in endotherms. Biological Reviews, 85, 545e569. Gilbert, C., Robertson, G., Le Maho, Y., & Ancel, A. (2008). How do weather conditions affect the huddling behaviour of emperor penguins? Polar Biology, 31, 163e169. Gilbert, C., Robertson, G., Le Maho, Y., Naito, Y., & Ancel, A. (2006). Huddling behavior in emperor penguins: dynamics of huddling. Physiology and Behavior, 88, 479e488. Hayes, L. D. (2000). To nest communally or not to nest communally: a review of rodent communal nesting and nursing. Animal Behaviour, 59, 677e688. Hothorn, T., Bretz, F., & Westfall, P. (2008). Simultaneous inference in general parametric models. Biometrical Journal, 50, 346e363. Kharitonov, S. P., & Siegel-Causey, D. (1988). Colony formation in seabirds. Current Ornithology, 5, 223e272.
98
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
Krause, J., & Ruxton, G. D. (2002). Living in groups. Oxford, U.K.: Oxford University Press. Le Bohec, C., Gauthier-Clerc, M., & Le Maho, Y. (2005). The adaptive significance of ches in the king penguin. Animal Behaviour, 70, 527e538. cre Le Maho, Y. (1977). The emperor penguin: a strategy to live and breed in the cold. American Scientist, 65, 680e693. Le Maho, Y., Delclitte, P., & Chatonnet, J. (1976). Thermoregulation in fasting emperor penguins under natural conditions. American Journal of Physiology, 231, 913e922. Le Maho, Y., Hanuise, N., Whittington, J. D., Pereira, L., Brucker, M., Chatelain, N., et al. (2014). Rovers to minimize human disturbance in research on wild animals. Nature Methods, 11, 1242e1244. Lewis, F., Butler, A., & Gilbert, L. (2011). A unified approach to model selection using the likelihood ratio test. Methods in Ecology and Evolution, 2, 155e162. Martin, R. A., Fiorentini, M., & Connors, F. (1980). Social facilitation of reduced oxygen consumption in Mus musculus and Meriones unguiculatus. Comparative Biochemistry and Physiology A, 65, 519e522. McCafferty, D. J., Gilbert, C., Thierry, A. M., Currie, J. C., Le Maho, Y., & Ancel, A. (2013). Emperor penguin body surfaces cool below air temperature. Biology Letters, 9, 20121192. cologiques ologie (Terre Mougin, J. L. (1966). Observations e a la colonie de Pointe Ge lie) en 1964. L'Oiseau et la Revue Française d'Ornithologie, 36, 166e226. Ade Ostner, J. (2002). Social thermoregulation in redfronted lemurs (Eulemur fulvus rufus). Folia Primatologica, 73, 175e180. Pierard, C., & Pettre, P. (1993). Some aspects of the climatology of Dumont d'Urville, lie Land, Antarctica. International Journal of Climatology, 13, 313e327. Ade Pinshow, B., Fedak, M. A., Battles, D. R., & Schmidt-Nielsen, K. (1976). Energy expenditure for thermoregulation and locomotion in emperor penguins. American Journal of Physiology, 231, 903e912. vost, J. (1961). Ecologie du manchot empereur Aptenodytes forsteri Gray. Paris, Pre France: Hermann. R Core Team. (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Robertson, G. (1990). Huddles. Australian Geographic, 20, 76e97. Siegel-Causey, D., & Kharitonov, S. P. (1991). The evolution of coloniality. Current Ornithology, 7, 285e330.
Waters, A., Blanchette, F., & Kim, A. D. (2012). Modeling huddling penguins. PLoS One, 7, e50277. Willis, C. K. R., & Brigham, R. M. (2007). Social thermoregulation exerts more influence than microclimate on forest roost preferences by a cavity-dwelling bat. Behavioral Ecology and Sociobiology, 62, 97e108. Zitterbart, D. P., Wienecke, B., Butler, J. P., & Fabry, B. (2011). Coordinated movements prevent jamming in an emperor penguin huddle. PLoS One, 6, e20260.
APPENDIX Table A1 Means ± SD (N) and ranges of climatic variables across reproductive seasons in 2008 Climatic variables
Air temperature ( C) Atmospheric pressure (hPa) Relative humidity (%) Wind speed (m/s) Wind direction ( ) Solar radiation (J/cm)
Reproductive season Pairing (3 Maye28 May)
Incubation (29 Maye15 July)
Chick rearing (16 Julye16 September)
13.3±3.8 (208) 22.9 to 6.2 988±11.4 (208) 956 to 1003 66.4±15.7 (208) 34.0 to 90.0 9.4±6.8 (208) 1.0 to 31.0 141±40.4 (208) 40.0 to 330 50.7±36.5 (26) 7 to 162
18.6±5.4 (384) 31.1 to 7.7 982±6.6 (384) 964 to 1003 49.3±14.4 (384) 16.0 to 85.0 7.1±6.3 (384) 0 to 35.0 150±59.1 (384) 0 to 360 10.4±8.7 (48) 2 to 45
15.5±4.6 (504) 31.5 to 1.2 983±12.8 (504) 951 to 1013 71.4±14.4 (504) 40.0 to 98.0 9.1±6.6 (504) 0 to 31.0 139±50.1 (504) 0 to 360 166±229 (63) 2 to 839
Contents lists available at ScienceDirect
Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav
New insights into the huddling dynamics of emperor penguins Ancel a, b, *, Caroline Gilbert c, Nicolas Poulin d, Michae €l Beaulieu e, Andre a, b Bernard Thierry Universit e de Strasbourg, Institut Pluridisciplinaire Hubert Curien, Strasbourg, France Centre National de la Recherche Scientifique, D epartement Ecologie, Physiologie et Ethologie, Strasbourg, France c Universit e Paris-Est, Ecole Nationale V et erinaire d'Alfort, Centre National de la Recherche Scientifique, Mus eum National d'Histoire Naturelle, Paris, France d CeStatS, Institut de Recherche Math ematique Avanc ee, Universit e de Strasbourg, Centre National de la Recherche Scientifique, Strasbourg, France e Zoological Institute and Museum, University of Greifswald, Greifswald, Germany a
b
a r t i c l e i n f o Article history: Received 13 March 2015 Initial acceptance 9 April 2015 Final acceptance 31 August 2015 Available online MS. number: 15-00215R Keywords: aggregation Antarctica Aptenodytes forsteri energy saving environment huddling social thermoregulation
Social thermoregulation is a cooperative strategy in which animals actively aggregate to benefit from the warmth of conspecifics in response to low ambient temperatures. Emperor penguins, Aptenodytes forsteri, use this behaviour to ensure their survival and reproduction during the Antarctic winter. An emperor penguin colony consists of a dynamic mosaic of compact zones, the so-called huddles, included in a looser network of individuals. To maximize energy savings, birds should adjust their huddling behaviour according to environmental conditions. Here, we examined the dynamics of emperor penguin aggregations, based on photo and video records, in relation to climatic factors. Environmental temperature, wind and solar radiation were the main factors contributing to huddle formation. The analysis of individual movements showed that birds originating from loose aggregations continually joined huddles. Sometimes, a small number of birds induced a movement that propagated to the entire huddle, causing its breakup within 2 min and releasing birds, which then integrated into looser aggregations. Different parts of the colony therefore appeared to continually exchange individuals in response to environmental conditions. A likely explanation is that individuals in need of warmth join huddles, whereas individuals seeking to dissipate heat break huddles apart. The regular growth and decay of huddles operates as pulses through which birds gain, conserve or lose heat. Originally proposed to account for reducing energy expenditure, the concept of social thermoregulation appears to cover a highly dynamic phenomenon that fulfils a genuine regulatory function in emperor penguins. © 2015 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Coloniality is the aggregation of individuals of the same species on a breeding ground remote from foraging sites (Kharitonov & Siegel-Causey, 1988). This kind of social reproduction is used by many species of vertebrates, and especially by seabirds among which more than 95% nest in colonies (Brown, Stutchbury, & Walsh, 1990). Living within a colony may entail costs in terms of intraspecific competition, spread of pathogens or travel distances (e.g. Kharitonov & Siegel-Causey, 1988; Krause & Ruxton, 2002). To be adaptive, coloniality has to bring fitness benefits that offset these zilly, & Pagel, 1988; costs (e.g. Danchin & Wagner, 1997; Dubois, Ce Siegel-Causey & Kharitonov, 1991). For some species, one main advantage of group living is social thermoregulation (Gilbert et al., 2010). This is a cooperative
* Correspondence: A. Ancel, CNRS/IPHC/DEPE, 23 rue Becquerel, 67087 Strasbourg, France. E-mail address: [email protected] (A. Ancel).
strategy that operates through both physiological and behavioural processes. It can be defined as the active aggregation of individuals to benefit from the warmth of conspecifics in response to low temperatures (Alberts, 1978; Gilbert et al., 2010; Martin, Fiorentini, & Connors, 1980). Huddling animals maximize energy savings by reducing the body surface area exposed to cold, thus decreasing heat loss, and by warming their surrounding environment. Huddling occurs in vertebrates as diverse as rodents, bats, primates and penguins (e.g. Gilbert, Blanc, Le Maho, & Ancel, 2008; Hayes, 2000; Ostner, 2002; Willis & Brigham, 2007). Huddles of emperor penches of king penguin chicks, guins, Aptenodytes forsteri, and cre Aptenodytes patagonicus, can involve thousands of birds, which cooperate to share warmth and decrease predation risk (Ancel, Visser, Handrich, & Le Maho, 1997; Gilbert, Robertson, Le Maho, Naito, & Ancel, 2006; Le Bohec, Gauthier-Clerc, & Le Maho, 2005; vost, 1961). Pre Huddling is probably a key factor allowing the colonization of particularly cold habitats by endotherms. Emperor penguins
http://dx.doi.org/10.1016/j.anbehav.2015.09.019 0003-3472/© 2015 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
92
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
represent the most extreme example, as they are the only animals able to breed during the Antarctic winter, far from the open sea vost, 1961). Their thermoneutral zone ranges where they feed (Pre from 10 to þ20 C, and they can withstand negative ambient temperatures only to a certain extent before increasing their metabolism to generate heat (Le Maho, Delclitte, & Chatonnet, 1976; Pinshow, Fedak, Battles, & Schmidt-Nielsen, 1976). However, when conditions deteriorate further (at 25 C, the probability of huddling is ca. 80% for a wind speed of 0 m/s and 98% for a wind speed of 15 m/s, Gilbert, Robertson, Le Maho, & Ancel, 2008), behavioural strategies complement physiological strategies, which allows emperor penguins to save energy (Le Maho, 1977). Under similar weather conditions, free-ranging males that huddled maintained a metabolic rate 16% lower than captive males that were experimentally prevented from forming dense aggregations (Ancel et al., 1997). When wind chill increases the cooling rate, emperor penguin vost, colonies move downwind (Gilbert, Robertson et al., 2008; Pre 1961; Robertson, 1990). By contrast, when wind chill decreases, colonies move against the wind, presumably because birds tend to return to their usual sites to avoid areas of unstable ice. Therefore, colonies move back and forth under the influence of winds (Gilbert, Robertson et al., 2008). It is known that a major part of energy saving in loosely grouped birds is from mutual wind protection (Ancel et al., 1997). Low ambient temperatures, however, appear more influential than wind speed in inducing the formation of huddles (Gilbert, Le Maho, Perret, & Ancel, 2007; Gilbert, Robertson et al., 2008), but we do not know to what extent other meteorological factors can affect the grouping patterns of emperor penguins. A colony of emperor penguins is spatially heterogeneous, consisting of several parts of varying densities (Gilbert et al., 2006). In the tighter aggregations called huddles, density can reach vost, 1961), individuals step in a synchronized 8e10 birds/m2 (Pre way (Gerum et al., 2013; Zitterbart, Wienecke, Butler, & Fabry, 2011) and the birds most exposed to the wind move along the opposite vost, 1961; flank of the huddle for protection (Birr, 1968; Pre Robertson, 1990; Waters, Blanchette, & Kim, 2012). As heat loss is limited within huddles (as heat can dissipate only by inhalation of cold air and through the head) ambient temperature can reach 37.5 C (Gilbert et al., 2006), well above the birds' þ20 C upper critical temperature (Le Maho et al., 1976; Pinshow et al., 1976). As a consequence, birds face the paradox that in a cold physical environment they sometimes need to dissipate excess heat (Gilbert et al., 2007, 2006). At present, we do not know how emperor penguins deal with these apparently contradictory requirements since the factors influencing the process of huddle formation and breakup have never been studied in detail with regard to environmental variables. Most studies have examined the effects of weather conditions on emperor penguin colony structure only during incubation, but since energetic constraints are likely to differ between the breeding stages it is important to examine them also during pairing and chick rearing. For instance, penguins have to fast for a prolonged time during incubation when weather conditions are most severe, while they only stay in the colony for short periods when rearing chicks, when weather conditions are more favourable. Consequently, forming huddles may be more critical during incubation relative to other breeding stages to maximize energy savings. To understand the dynamics of social thermoregulation in emperor penguins according to environmental variables (ambient temperature, wind speed, wind direction, relative humidity, atmospheric pressure or solar radiation), we investigated the huddling behaviour of an entire colony during pair formation, incubation and chick rearing, that is, from winter to summer. We
aimed to test the following predictions: (1) as dense aggregations protect emperor penguins from inclement weather conditions and help individuals to save energy, meteorological variables that have the potential to lower body temperature should result in an increase in huddle formation and duration, (2) as social thermoregulation is based upon the need of individuals to save energy by forming huddles and the need to shed excess heat by leaving huddles, this should lead to the regular formation and breakup of huddles, and (3) as a colony of emperor penguins is made of aggregations of different densities, individuals should move between loose aggregations and huddles to control body temperature. METHODS Subjects ologie We studied the emperor penguin colony of Pointe Ge lie, Antarctica, 66 400 S, Archipelago (Dumont d'Urville, Terre Ade 140 010 E) where about 3000 pairs breed each year on the fast-ice vost, 1961). We between islands and the Antarctic continent (Pre collected data in 2005, 2006 and 2008, over three reproductive periods: pairing (from arrival of the birds in late April to mid-May), incubation (from mid-May to mid-July) and chick rearing (from mid-July to mid-September). Fieldworkers counted the birds in the colony at each season, once a week depending on the weather conditions (good visibility, low wind). There were a maximum of 3253, 3350 and 3160 breeding males during the incubation periods in 2005, 2006 and 2008, respectively. Definitions To analyse the distribution of emperor penguins within the colony, we differentiated between two different grouping patterns: huddles and loose aggregations. A huddle was defined as a part of a group in which individuals were closely assembled; typically, individuals had flippers held against the body and the head tucked into the shoulders, and each bird hid its beak between the necks of others. Note that movements within a huddle are limited, birds being only able to move the head or to make limited steps in a synchronized way (Gerum et al., 2013; Zitterbart et al., 2011). Sometimes a huddle broke down when its members became more active (Fig. 1). A group could contain one or several huddles. Loose aggregations consisted of all the individuals in a group that did not belong to huddles. These birds were stationary or mobile, and had limited if any physical contact with their neighbours or were apart from each other. They engaged in different activities such as grooming, lifting the abdominal skin fold covering the egg, consuming snow or performing courtship displays when females were present. During the pairing period, both males and females were present in the huddles but during the incubation period, only males were present. Climatic and Seasonal Influences on Huddling Patterns To investigate huddling patterns, we photographed the colony every day (weather permitting) from late April to mid-September 2008. Several photos were taken within three different time windows (morning: 0930e1130 hours; midday: 1130e1330 hours; afternoon: 1330e1530 hours; local time i.e. UTC þ 10 h), each time from three vantage points located on nearby hills (from two points on Le Mauguen Island and from one point on Rostand Island). The different photographs of the colony were collated to get a single picture of the entire colony using PTGui software (www.ptgui.com). Each picture was visually analysed on a computer screen using the zoom function. This led to a database of 118 panoramic pictures: 20
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
93
Figure 1. Grouping patterns in emperor penguins during the incubation period: (a) huddles on the centre and bottom of the picture, with loose aggregations around them; (b) the huddle in the centre starts to break on the right side; (c) after breakup birds reach a state of loose aggregation.
during pairing (morning: 11; midday: 0; afternoon: 9), 60 during incubation (21, 21, 18) and 38 during chick rearing (14, 11, 13). We obtained meteorological data from the permanent weather te o France at Dumont d'Urville located on Petrel station of Me
Island, at 43 m in height and 500 m from the colony of emperor penguins. Because temperatures and wind speed have already been described as being lower at the colony site than at the meteorological station, they were corrected for the purpose of our study
94
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
according to the equations given by Gilbert et al. (2007). We calculated a daily mean from measures taken every 3 h: air temperature, atmospheric pressure, relative humidity at the corresponding air temperature, wind speed and wind direction; solar radiation was summed on a daily basis (Appendix Table A1). Duration of Huddles, Breakups and Movements of Individuals
huddles and mean number of individuals per huddle. Predictor variables were meteorological factors (see above), reproductive and daytime periods, and their interactions. We considered only simple interactions since the models did not converge with multiple interactions. Each quantitative predictor variable was scaled before being used in the models. As the number of huddles corresponded to count data, generalized linear models (GLM) using a quasiPoisson distribution were fitted. When using the quasi-Poisson family, the Akaike information criterion is not available; hence we used likelihood ratio tests (LRT) to select the best model (Lewis, Butler, & Gilbert, 2011). To obtain P values for each variable, only interactions were removed according to LRT. We applied this procedure for all models. With regard to the mean number of individuals per huddle, residuals followed a normal law when data were log-transformed, so we used a linear model on the logtransformed mean number of individuals per huddle. For the logtransformed mean number of individuals per huddle, the best model was selected using LRT. Post hoc comparisons were made using the R package ‘multcomp’ with Tukey corrections (Hothorn, Bretz, & Westfall, 2008). We analysed the movements of individuals between huddles and loose aggregations using the Wilcoxon signed-ranks test, exact procedure. The significance level was set at 0.05. Average values are given as means and standard errors of the mean (SEM).
To investigate the dynamics of huddles and the movements of individuals during daylight, we videotaped the colony during the incubation period using a digital camera between 1000 and 1600 hours local time, i.e. from sunrise to sunset. We collected 22 h of recording over 7 days in 2005 (from 6 to 12 July) and 46 h over 16 days in 2006 (between 25 May and 28 June). Images were visually analysed frame by frame. A bird was considered as moving when it advanced at least three steps. It could walk, stop or join a huddle. It could also leave a huddle and reach a loose aggregation. We estimated the duration of huddles from six video clips, recorded in June 2006, which displayed the entire colony for at least 3 h. We counted the huddles at 15 min intervals (N ¼ 13) using scan sampling. We also counted the breakups occurring across the whole duration of videos. The mean duration of huddles was defined as the ratio of the total number of huddles over all scans in the colony divided by the number of breakups, thus providing measures given in number of 15 min intervals. We assessed the time individuals spent in huddles from data recorded with data loggers (Mk9, Wildlife Computers, Redmond, WA, U.S.A.) deployed on 13 males in 2005 and 2006. The tags, which were glued to their lower back, recorded ambient temperature and light intensity every 5 s in 2005 and every 2 s in 2006 (Ancel, Beaulieu, Le Maho, & Gilbert, 2009). An emperor penguin was deemed to be in a huddle when its back was completely covered by another bird behind it and the recorded light intensity dropped below 40 arbitrary units (a.u.); light exceeded 120 a.u. during the day and was 60e80 a.u. at night. When a bird entered a huddle the temperature increased at the same time as the light level dropped and vice versa when it left the huddle (see Ancel et al., 2009). We examined the origin of breakups from 20 huddles for which the start of breakup was videotaped. We assessed the temporal dynamics of breakups from nine video clips in which the whole process was recorded. We then selected six breakups for which we could follow individuals during the 5 min following breaking up in order to investigate the outcome of breakups. To assess the temporal dynamics of breakups, we recorded for every bird belonging to the huddle whether they were motionless or showing any sign of activity at 20 s intervals. We assessed the expansion of a huddle during a breakup by measuring the maximal length and width of the area occupied by the birds before and after breakup. To estimate the area covered by a huddle we assumed that it was shaped like an ellipse. In a final step, we determined the time elapsed between a breakup and the subsequent re-establishment of a huddle (i.e. at least 20 individuals) in the same location by monitoring the outcomes of 23 breakups for as long as allowed by video records. To study the movements of individuals between huddles and loose aggregations, we randomly selected 11 sequences from five video clips that displayed both a loose aggregation and more than one huddle, with no breakup occurring during these sequences. We counted the birds in huddles and loose aggregations, and the number of these moving from one type of grouping to another.
The number of huddles increased when air temperature and wind speed diminished and when solar radiation rose and also changed according to wind direction (Table 1). We additionally found a negative interaction between wind direction and relative humidity (Table 1). The mean number of individuals per huddle increased when air temperature or solar radiation decreased, and when wind speed increased (Table 1). The reproductive period had a significant effect on the two target variables (Table 1). Significant interactions appeared between reproductive season and climatic factors: with air temperature for number of huddles and with solar radiation for mean number of individuals per huddle (Table 1). Additionally, an interaction appeared between daytime period and wind direction for the number of huddles (Table 1). Huddling patterns by periods are presented in Table 2. The two target variables varied as a function of reproductive period. There were fewer huddles during the incubation period (3.6 ± 1.2) than during the pairing (6.5 ± 1.3) and chick-rearing periods (11.8 ± 2.1; Tukey tests: pairing versus incubation: P < 0.001; incubation versus chick rearing: P < 0.001; pairing versus chick rearing: P ¼ 0.144). The mean number of individuals per huddle was higher during the incubation period (897 ± 35) than the chick-rearing period (123 ± 24; Tukey test: P ¼ 0.013), but no significant differences appeared between pairing (246 ± 87) and incubation (P ¼ 0.256) or chick-rearing periods (P ¼ 0.630).
Statistical Analyses
Duration and Breakup of Huddles
We analysed daily data on climatic and seasonal influences with R 2.15.0 (R Core Team, 2012). Target variables were number of
Birds spent on average 62.1 ± 2.9 min (N ¼ 50 huddles) and 41.0 ± 0.6 min (N ¼ 185 huddles) in huddles in 2005 (7 days) and
Ethical Note The study was approved by the ethics committee of the Institut Polaire Français Paul-Emile Victor (IPEV) and the Terres Australes et Antarctiques Françaises (TAAF). RESULTS Climatic and Seasonal Influences on Huddling Patterns
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
95
Table 1 Results of GLM analyses testing the influence of climatic factors, reproductive and daytime periods and their interactions on huddling patterns Predictor variable
Number of huddles
Air temperature Atmospheric pressure Relative humidity Solar radiation Wind speed Wind direction Reproductive period Daytime period Air temperature ) reproductive period Wind direction ) relative humidity Wind direction ) daytime period Solar radiation ) reproductive period
Mean number of individuals per huddle
Adjusted generalized VIF
Estimate
Decision statistics (t or F)
P
Adjusted generalized VIF
Estimate
Decision statistics (t or F)
P
2.13 1.41 1.37 1.43 1.39 2.44 1.82 1.07 1.82 1.76 3.90
0.348 0.118 0.006 0.141 0.172 0.438
2.80 1.74 0.08 2.02 2.10 3.94 30.30 0.99 19.4 3.75 15.4
0.006 0.085 0.935 0.046 0.039 <0.001 <0.001 0.374 <0.001 <0.001 <0.001
1.76 1.38 1.30 2.75 1.40 1.32 7.63 1.05
0.472 0.115 0.036 0.939 0.416 0.188
3.38 1.07 0.36 4.34 3.86 1.80 24.2 0.894
0.001 0.288 0.721 <0.001 <0.001 0.074 <0.001 0.412
14.30
<0.001
0.305
2.20
VIF ¼ variance inflation factor. Estimates and t values are provided for the effects of climatic factors and F values for the effects of periods. Significant results are indicated in bold.
Table 2 Huddling patterns according to reproductive and daytime periods (mean ± SEM) Huddling patterns
Pairing
Number of huddles Number of individuals per huddle
Incubation
Chick rearing
Morning (11)
Midday (0)
Afternoon (9)
Morning (21)
Midday (21)
Afternoon (18)
Morning (14)
Midday (11)
Afternoon (13)
6.2±1.7 262±87
e e
6.8±0.9 226±88
4.3±1.4 798±39
3.5±1.3 1106±38
3.0±0.9 769±26
12.7±2.1 186±16
13.5±2.6 103±27
9.4±1.6 73±29
Number of days is given in parentheses.
Percentage of moving individuals
100
80
60
40
20
0 –160 –120
–80
–40 0 40 80 20 s time intervals
120
160
200
Figure 2. Dynamics of the breakup process: mean percentage of moving individuals over time (at time interval 0 the number of moving individuals passes from 50 to over 50).
number of 599 ± 243 birds). Most dispersing birds joined loose aggregations (99.7 ± 0.2% of individuals; Fig. 1c) rather than other huddles (0.3 ± 0.2%). Huddles did not form again immediately after breakup; they re-established at the same location within 1 min in only 12% of cases, while after 40 min 80% of huddles had reformed (Fig. 3).
Movements of Individuals Between Huddles and Loose Aggregations The percentage of individuals transferring from a loose aggregation to a huddle (2.1 ± 0.6% of individuals staying in loose aggregations per min) was significantly higher than the percentage of individuals leaving a huddle to join a loose aggregation (0.3 ± 0.1%
Cumulative percentage of reformed huddles
2006 (16 days), respectively. Huddles could last less than one, and more than 13, 15 min intervals (i.e. more than ca. 3 h). We found a mean duration of 6.1 ± 1.9 min at 15 min intervals per huddle, i.e. about 90 min. Over the 20 huddles for which the start of the breakup was videotaped, 19 started near the edge of the huddle (see Fig. 1b) and one from the centre. One breakup was induced by a conflict between two individuals, six by the departure of several individuals, and no visible cause was detected for the remaining 13 events. During breakup the percentage of moving individuals rose from 2.7 ± 0.7% to 99.9 ± 0.1% within 2 min for a mean number of 434 ± 88 birds (Fig. 2). With regard to spatial dynamics, the ratio between the area occupied by individuals in a huddle and that occupied by these individuals after breakup was 1.8 ± 0.1 (mean
100
80
60
40
20
0
1
5
10 20 30 40 60 80 100 120 140 160 Time (min)
Figure 3. Latency between breakup and re-establishment of huddling in the same location of the colony (N ¼ 23 huddles).
96
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
of individuals staying in loose aggregations per min; Wilcoxon test: N ¼ 11 sequences, z ¼ 2.76, P ¼ 0.003). Focusing on huddles, we then determined from where individuals came when aggregating to huddle, or where they went when they left a huddle. We compared the flow of individuals entering and leaving a huddle within 1 min. The percentage of birds entering the huddle was three times as large as the percentage of individuals leaving it (75.0 ± 4.7% and 25.0 ± 4.7% per min, respectively; Wilcoxon test: N ¼ 22 huddles, z ¼ 3.47, P < 0.001). Among those integrating into a huddle, the percentage of birds coming from a loose aggregation was 10 times as large as the percentage of birds coming from a huddle (90.9 ± 8.3% and 9.1 ± 8.3% per min, respectively; Wilcoxon test: N ¼ 8 huddles, z ¼ 2.52, P ¼ 0.008). Among those leaving a huddle, the percentage of birds entering a loose aggregation was four times as large as the percentage of birds joining a huddle (80.0 ± 4.8% and 20.0 ± 4.8% per min, respectively; Wilcoxon test: N ¼ 8 huddles, z ¼ 2.52, P ¼ 0.008). To test whether the transfer of penguins was influenced by wind direction, we compared for 21 huddles the percentages of birds aggregating on the side of the huddle exposed to the wind versus the leeward side (62.4 ± 5.3% and 37.6 ± 5.3%, respectively). This did not produce a statistically significant difference (Wilcoxon test: N ¼ 21, z ¼ 1.40, P ¼ 0.168). No significant differences emerged either when comparing the percentages of birds leaving the exposed side versus the leeward side (36.6 ± 8.2% and 63.4 ± 8.2%, respectively; Wilcoxon test: N ¼ 12, z ¼ 1.36, P ¼ 0.203). DISCUSSION Overall, our results show that air temperature, wind and solar radiation were the main drivers pushing emperor penguins to gather in huddles. The influence of weather conditions on huddle dynamics was reflected by variable huddling patterns across breeding stages. Moreover, individuals regularly shifted between aggregations of different densities; they slowly moved from loose aggregations to huddles, while they rapidly left huddles following breakups. Climatic and Seasonal Influences on Huddling Patterns The number of huddles and the number of individuals per huddle increased when air temperature decreased. Moreover, when wind speed rose or solar radiation diminished, the mean number of individuals per huddle increased, while the number of huddles decreased, meaning that huddles got larger. Other climatic factors such as wind direction and relative humidity also favoured the formation of huddles. The effects of air temperature on the number of huddles is surprising, as one would expect a few large huddles to enhance heat conservation when the temperature sinks. One reason may be that when temperatures are low, penguins form huddles with the closest individuals (they do not look for a large pre-existing huddle), so that a large number of small huddles form. In contrast, when conditions are milder, penguins may take time to find a pre-existing huddle. The breeding stages corresponded to different physiological and demographic conditions. In the pairing period, females were still present and the colony remained dispersed. During incubation, males had to minimize heat loss to survive fasting while incubating vost, 1961). During chick rearing, chicks (Gilbert et al., 2006; Pre were protected against cold and predation, and were fed alternately by the parents. Moreover, the pairing, incubation and chick-rearing periods corresponded to different climatic seasons (Pierard & Pettre, 1993). In our study, the number of huddles decreased during incubation relative to the pairing and the chick-rearing periods.
In parallel, the largest number of penguins per huddle was found during incubation (up to 1100 individuals) while during chick rearing the number was lowest (fewer than 200 birds). As a corollary, loose aggregations were more frequent during chick rearing than at other periods. It may be hypothesized that this seasonal huddling pattern reflects the number of individuals present in the colony. Indeed, during the pairing period, males and females are present in the colony while only males are present during incubation, and adults are sporadically present at the end of the chickrearing period. The number of huddles would therefore be expected to be about twice as large during the pairing period as during incubation, and lower during incubation than during the chick-rearing period. Although this expected difference was observed between the pairing period and incubation, it was not found between the chick-rearing period and other periods. This indicates that factors other than the number of birds influence huddling patterns in emperor penguins. Accordingly, we still found significant effects of air temperature, solar radiation, wind speed and wind direction on huddling patterns after controlling for reproductive season. Duration and Breakup of Huddles The average time spent by individuals in a huddle fluctuated around 50 min, and huddles themselves lasted between a dozen of minutes to several hours. These numbers compare well with a huddle duration of 1.6 ± 1.7 h (mean ± SD), as previously reported in the same colony (Gilbert et al., 2006). These results indicate that emperor penguins packed in huddles remain in this state only for relatively short periods. By following the fate of huddles, we were able to highlight their breaking-up processes. Breakups could originate from the departure of some individuals or from agonistic behaviours between neighbours, but in most cases no visible cause was found (see also Mougin, 1966; Robertson, 1990). In accordance with our second prediction, breakups appeared as a regular process through which emperor penguins collectively left huddles. Breakups generally started close to the edge of the huddle, then propagated within 2 min to the entire huddle. Similarly, waves moving across the huddle appear to be initiated by a few individuals (Gerum et al., 2013). According to the thermoregulatory hypothesis, breakups were expected to occur more frequently in the middle of huddles, where temperature may be maximal. However, our results indicate that this may not be the case, which suggests that ambient temperatures may not necessarily increase linearly from the edge to the centre of the huddle. Another explanation is that penguins in the centre are stuck in the huddle, while penguins at the periphery can move enough to initiate huddle breakups more frequently. Breakups expanded until the huddles reached an area almost twice as large as that they initially occupied. Within 5 min after breakup, the previous members of a huddle seldom joined another huddle, meaning that almost all of them then belonged to the loosely aggregated part of the colony. Breakups therefore appear as the very process by which emperor penguins can release themselves from mutual body contact and dissipate excess heat. The breakup of huddles is sometimes accompanied by a haze of warm air rising over the colony, which indicates a significant dissipation of heat (Robertson, 1990). A previous study showed that penguins leave huddles when their body temperature reaches 37.5 C (Ancel et al., 2009). Those penguins whose skin temperature (37.5 C) is close to the animal's deep body temperature (Gilbert et al., 2006) could aim to find a lower external temperature to cool down. Related to this, we also observed that some birds leaving a huddle consumed fresh snow. Finally, other motives unrelated to thermoregulation may lead birds to initiate a breaking-up process, such
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
as the need to move freely or avoid agonistic behaviours within the huddle. Another important factor that may lead penguins to leave the huddle is that incubating birds need to regularly turn their egg to ensure the proper development of the embryo and thereby hatching success (Elibol & Brake, 2004). Movements of Individuals Between Huddles and Loose Aggregations The analysis of individual movements showed that most movements were directed from loose aggregations to huddles. This indicates that huddles were growing when observations were conducted. Individuals continually percolate from loose aggregations to huddles, while they join loose aggregations through the abrupt breaking up of huddles. To our knowledge, no equivalent process has ever been reported in another species, but it might also che period (Le Bohec occur in king penguin chicks during the cre et al., 2005). It is worth noting that some emperor penguins located on the outer edge of huddles can leave them without disturbing their neighbours. These birds also preferred to integrate into loose aggregations rather than join other huddles, which is consistent with the thermoregulatory hypothesis. If animals look for colder conditions, it is understandable that they do not immediately join another huddle but need time outside huddles to dissipate heat. It might happen that birds located on the outer edge of huddles need to reduce their exposure to the cold. Some authors have proposed that birds that are most exposed to the wind regularly break away to reach the sheltered side of the huddle (Birr, 1968; vost, 1961; Robertson, 1990). However, we did not find eviPre dence for such a phenomenon in our study. The birds did not leave the leeward side more often than the exposed side, and they did not show any significant preference for one side or the other when joining a huddle. The number of departures from huddles was negligible compared to the number of arrivals. When a bird integrates into a huddle, it is quickly surrounded by other individuals aggregating behind it, suggesting that it probably does not need to choose a particular side. Conclusion and Perspectives In agreement with our third expectation, our results support the view that emperor penguin colonies represent dynamic networks, the different parts of which continually exchange individuals in response to environmental conditions (Gilbert et al., 2006). It appears that individuals in need of warmth join huddles, and loose aggregations continually feed huddles. Conversely, huddles feed loose aggregations back, albeit in a discontinuous way. This temporal difference between joining and leaving a huddle presumably comes from the fact that individuals in loose aggregations differ in their need for warmth, while individuals in huddles simultaneously need to dissipate heat. The regular growth and decay of huddles therefore appear as pulses through which birds conserve or lose heat. Originally proposed to account for the reduction in energy expenditure made possible by social grouping, the concept of social thermoregulation appears to cover a dynamic phenomenon that fulfils a genuine regulatory function in emperor penguins. We acknowledge that our results hold for one colony of emperor lie, and caution is needed when generalizing penguins in Terre Ade to other colonies. It would be necessary to verify that our results are not due to particular meteorological conditions, and that similar processes occur in other sites too. For instance, the size of the ologie compared to other colony is relatively small at Pointe Ge locations. Some colonies are even smaller, however, making social thermoregulation more difficult. How the size of the colony can affect huddling patterns in emperor penguins remains to be
97
examined. Moreover, it should be kept in mind that our visual records necessarily took place during periods of good visibility, that is, under relatively good weather conditions, and when the safety rules of the French station permitted fieldwork. Because of low visibility, we could not make records in the most critical climatic conditions faced by the birds, and in particular at night or during severe blizzards for evident safety reasons. The next step will be to use covered infrared cameras to continuously monitor huddle formation and breakup during blizzards and at night, by time-lapse recording. This may be complemented by the deployment of operated stations such as wireless networks or rovers (Le Maho et al., 2014), able to identify an animal previously equipped with a transponder and to record its position accurately by using a GPS. Finally, developing software able to automatically follow the same individuals for long periods (Gerum et al., 2013; McCafferty et al., 2013; Zitterbart et al., 2011) offers the opportunity to improve analyses of huddling dynamics in colonies of emperor penguins. Acknowledgments Fieldwork was financially and logistically supported by the Institut Polaire Français Paul-Emile Victor (IPEV) and the Terres te o France generously Australes et Antarctiques Françaises. Me provided meteorological data. We thank Antoine Dervaux and le ne Petit and Vincent Anne-Mathilde Thierry for fieldwork, and He Martin for data analysis. Finally, we particularly thank the two anonymous referees who provided valuable comments and contributed to a significant improvement of this paper. References Alberts, J. R. (1978). Huddling by rat pups: group behavioral mechanisms of temperature regulation and energy conservation. Journal of Comparative and Physiological Psychology, 92, 231e245. Ancel, A., Beaulieu, M., Le Maho, Y., & Gilbert, C. (2009). Emperor penguins mates: keeping together in the crowd. Proceedings of the Royal Society B: Biological Sciences, 276, 2163e2169. Ancel, A., Visser, H., Handrich, Y., & Le Maho, Y. (1997). Energy saving in huddling penguins. Nature, 385, 304e305. tho-e cologiques a la colonie de manchots empereurs Birr, G. (1968). Observations e ologie en 1966. L'Oiseau et la Revue Française d'Ornithologie, 38, de Pointe Ge 53e88. Brown, C. R., Stutchbury, B. J., & Walsh, P. D. (1990). Choice of colony size in birds. Trends in Ecology & Evolution, 5, 398e403. Danchin, E., & Wagner, R. H. (1997). The evolution of coloniality: the emergence of new perspectives. Trends in Ecology and Evolution, 12, 342e347. zilly, F., & Pagel, M. (1988). Mate fidelity and coloniality in waterbirds: Dubois, F., Ce a comparative analysis. Oecologia, 116, 433e440. Elibol, O., & Brake, J. (2004). Identification of critical periods for turning broiler hatching eggs during incubation. British Poultry Science, 45, 631e637. Gerum, R. C., Fabry, B., Metzner, C., Beaulieu, M., Ancel, A., & Zitterbart, D. P. (2013). The origin of traveling waves in an emperor penguin huddle. New Journal of Physics, 15. njp483133. Gilbert, C., Blanc, S., Le Maho, Y., & Ancel, A. (2008). Energy saving processes in huddling emperor penguins: from experiments to theory. Journal of Experimental Biology, 211, 1e8. Gilbert, C., Le Maho, Y., Perret, M., & Ancel, A. (2007). Body temperature changes induced by huddling in breeding male emperor penguins. American Journal of Physiology, 292, R176eR185. Gilbert, C., McCafferty, D., Le Maho, Y., Martrette, J. M., Giroud, S., Blanc, S., et al. (2010). One for all and all for one: the energetic benefits of huddling in endotherms. Biological Reviews, 85, 545e569. Gilbert, C., Robertson, G., Le Maho, Y., & Ancel, A. (2008). How do weather conditions affect the huddling behaviour of emperor penguins? Polar Biology, 31, 163e169. Gilbert, C., Robertson, G., Le Maho, Y., Naito, Y., & Ancel, A. (2006). Huddling behavior in emperor penguins: dynamics of huddling. Physiology and Behavior, 88, 479e488. Hayes, L. D. (2000). To nest communally or not to nest communally: a review of rodent communal nesting and nursing. Animal Behaviour, 59, 677e688. Hothorn, T., Bretz, F., & Westfall, P. (2008). Simultaneous inference in general parametric models. Biometrical Journal, 50, 346e363. Kharitonov, S. P., & Siegel-Causey, D. (1988). Colony formation in seabirds. Current Ornithology, 5, 223e272.
98
A. Ancel et al. / Animal Behaviour 110 (2015) 91e98
Krause, J., & Ruxton, G. D. (2002). Living in groups. Oxford, U.K.: Oxford University Press. Le Bohec, C., Gauthier-Clerc, M., & Le Maho, Y. (2005). The adaptive significance of ches in the king penguin. Animal Behaviour, 70, 527e538. cre Le Maho, Y. (1977). The emperor penguin: a strategy to live and breed in the cold. American Scientist, 65, 680e693. Le Maho, Y., Delclitte, P., & Chatonnet, J. (1976). Thermoregulation in fasting emperor penguins under natural conditions. American Journal of Physiology, 231, 913e922. Le Maho, Y., Hanuise, N., Whittington, J. D., Pereira, L., Brucker, M., Chatelain, N., et al. (2014). Rovers to minimize human disturbance in research on wild animals. Nature Methods, 11, 1242e1244. Lewis, F., Butler, A., & Gilbert, L. (2011). A unified approach to model selection using the likelihood ratio test. Methods in Ecology and Evolution, 2, 155e162. Martin, R. A., Fiorentini, M., & Connors, F. (1980). Social facilitation of reduced oxygen consumption in Mus musculus and Meriones unguiculatus. Comparative Biochemistry and Physiology A, 65, 519e522. McCafferty, D. J., Gilbert, C., Thierry, A. M., Currie, J. C., Le Maho, Y., & Ancel, A. (2013). Emperor penguin body surfaces cool below air temperature. Biology Letters, 9, 20121192. cologiques ologie (Terre Mougin, J. L. (1966). Observations e a la colonie de Pointe Ge lie) en 1964. L'Oiseau et la Revue Française d'Ornithologie, 36, 166e226. Ade Ostner, J. (2002). Social thermoregulation in redfronted lemurs (Eulemur fulvus rufus). Folia Primatologica, 73, 175e180. Pierard, C., & Pettre, P. (1993). Some aspects of the climatology of Dumont d'Urville, lie Land, Antarctica. International Journal of Climatology, 13, 313e327. Ade Pinshow, B., Fedak, M. A., Battles, D. R., & Schmidt-Nielsen, K. (1976). Energy expenditure for thermoregulation and locomotion in emperor penguins. American Journal of Physiology, 231, 903e912. vost, J. (1961). Ecologie du manchot empereur Aptenodytes forsteri Gray. Paris, Pre France: Hermann. R Core Team. (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Robertson, G. (1990). Huddles. Australian Geographic, 20, 76e97. Siegel-Causey, D., & Kharitonov, S. P. (1991). The evolution of coloniality. Current Ornithology, 7, 285e330.
Waters, A., Blanchette, F., & Kim, A. D. (2012). Modeling huddling penguins. PLoS One, 7, e50277. Willis, C. K. R., & Brigham, R. M. (2007). Social thermoregulation exerts more influence than microclimate on forest roost preferences by a cavity-dwelling bat. Behavioral Ecology and Sociobiology, 62, 97e108. Zitterbart, D. P., Wienecke, B., Butler, J. P., & Fabry, B. (2011). Coordinated movements prevent jamming in an emperor penguin huddle. PLoS One, 6, e20260.
APPENDIX Table A1 Means ± SD (N) and ranges of climatic variables across reproductive seasons in 2008 Climatic variables
Air temperature ( C) Atmospheric pressure (hPa) Relative humidity (%) Wind speed (m/s) Wind direction ( ) Solar radiation (J/cm)
Reproductive season Pairing (3 Maye28 May)
Incubation (29 Maye15 July)
Chick rearing (16 Julye16 September)
13.3±3.8 (208) 22.9 to 6.2 988±11.4 (208) 956 to 1003 66.4±15.7 (208) 34.0 to 90.0 9.4±6.8 (208) 1.0 to 31.0 141±40.4 (208) 40.0 to 330 50.7±36.5 (26) 7 to 162
18.6±5.4 (384) 31.1 to 7.7 982±6.6 (384) 964 to 1003 49.3±14.4 (384) 16.0 to 85.0 7.1±6.3 (384) 0 to 35.0 150±59.1 (384) 0 to 360 10.4±8.7 (48) 2 to 45
15.5±4.6 (504) 31.5 to 1.2 983±12.8 (504) 951 to 1013 71.4±14.4 (504) 40.0 to 98.0 9.1±6.6 (504) 0 to 31.0 139±50.1 (504) 0 to 360 166±229 (63) 2 to 839