Combined use of two supervised learning algorithms to model sea turtle behaviours from tri-axial acceleration data

Accelerometers are becoming ever more important sensors in animal-attached technology, providing data that allow determination of body posture and movement and thereby helping to elucidate behaviour in animals that are difficult to observe. We sought to validate the identification of sea turtle behaviours from accelerometer signals by deploying tags on the carapace of a juvenile loggerhead (Caretta caretta), an adult hawksbill (Eretmochelys imbricata) and an adult green turtle (Chelonia mydas) at Aquarium La Rochelle, France. We recorded tri-axial acceleration at 50 Hz for each species for a full day while two fixed cameras recorded their behaviours. We identified behaviours from the acceleration data using two different supervised learning algorithms, Random Forest and Classification And Regression Tree (CART), treating the data from the adult animals as separate from the juvenile data. We achieved a global accuracy of 81.30% for the adult hawksbill and green turtle CART model and 71.63% for the juvenile loggerhead, identifying 10 and 12 different behaviours, respectively. Equivalent figures were 86.96% for the adult hawksbill and green turtle Random Forest model and 79.49% for the juvenile loggerhead, for the same behaviours. The use of Random Forest combined with CART algorithms allowed us to understand the decision rules implicated in behaviour discrimination, and thus remove or group together some 'confused' or under--represented behaviours in order to get the most accurate models. This study is the first to validate accelerometer data to identify turtle behaviours and the approach can now be tested on other captive sea turtle species.

Influence of weather conditions on the flight of migrating black storks

This study tested the potential influence of meteorological parameters (temperature, humidity, wind direction, thermal convection) on different migration characteristics (namely flight speed, altitude and direction and daily distance) in 16 black storks (Ciconia nigra). The birds were tracked by satellite during their entire autumnal and spring migration, from 1998 to 2006. Our data reveal that during their 27-day-long migration between Europe and Africa (mean distance of 4100 km), the periods of maximum flight activity corresponded to periods of maximum thermal energy, underlining the importance of atmospheric thermal convection in the migratory flight of the black stork. In some cases, tailwind was recorded at the same altitude and position as the birds, and was associated with a significant rise in flight speed, but wind often produced a side azimuth along the birds' migratory route. Whatever the season, the distance travelled daily was on average shorter in Europe than in Africa, with values of 200 and 270 km d(-1), respectively. The fastest instantaneous flight speeds of up to 112 km h(-1) were also observed above Africa. This observation confirms the hypothesis of thermal-dependant flight behaviour, and also reveals differences in flight costs between Europe and Africa. Furthermore, differences in food availability, a crucial factor for black storks during their flight between Europe and Africa, may also contribute to the above-mentioned shift in daily flight speeds.

Heart rate as a predictor of energy expenditure in undisturbed fasting and incubating penguins

Heart rate (f(H)) measurement offers the possibility to monitor energy expenditure (EE) in wild animals if the EE/f(H) relationship for the species, physiological stages and activities of interest is known. This relationship has been extensively studied using oxygen consumption rate ( ) measurement in captive, repeatedly handled king penguins (Aptenodytes patagonicus). Unfortunately, the potential effects of stress on the observed relationships resulting from handling and confinement were not considered. This study is the first involving undisturbed animals, and determines the EE/f(H) relationship in naturally fasting and freely incubating or captivity-acclimatized male and female king penguins. EE determination was based on (1) the measurement of body mass loss during periods of phase II fasting, and (2) the calculation of its energy equivalent from changes in body composition, i.e. 23.9 kJ g(-1). f(H) levels in freely incubating and captivity-acclimatized birds were found to be 50-70% lower than those previously reported for resting king penguins during measurements. Significant EE/f(H) relationships were found in freely incubating and captive males and females (R(2)=0.59 to 0.84), with no difference observed between genders. The best overall relationship was obtained by including fasting duration (t, days) in the model: EE=818+43.7xf(H)+36.3t-1.4txf(H) (R(2)=0.91). This equation yielded EE estimates approximately 26% higher than the previously reported 'best' predictive equation in king penguins, and even more so when f(H) was low. This result suggests that stress induces a disproportionate increase of f(H) vs O(2) consumption, and that the use of EE/f(H) relationships obtained in stressed birds could lead to underestimated EE values.

Recovery from swimming-induced hypothermia in king penguins: effects of nutritional condition

We investigated changes in the rate of oxygen consumption (V O2) and body temperature of wild king penguins (Aptenodytes patagonicus) in different nutritional conditions during recovery after exposure to cold water. Over time, birds undertook an identical experiment three times, each characterized by different nutritional conditions: (1) having recently completed a foraging trip, (2) after fasting for many days, and (3) having been refed one meal after the fast. The experiments consisted of a 2-h session in a water channel followed by a period of recovery in a respirometer chamber on land. Refed birds recovered significantly more quickly than fed birds, in terms of both time to reach resting V O2 on land and time to reach recovery of lower abdominal temperature. Previous work found that when penguins are in cold water, abdominal temperatures decrease less in refed birds than in fed or fasted birds, suggesting that refed birds may be vasoconstricting the periphery while perfusing the gut region to access nutrients. This, alongside an increased resting [V O2], seems the most reasonable explanation for why refed birds recovered more quickly subsequent to cold-water exposure in this study; that is, vasoconstriction of the insulative periphery meant that they lost less heat generated by the body core.

To what extent might N2 limit dive performance in king penguins?

A mathematical model was used to explore if elevated levels of N2, and risk of decompression sickness (DCS), could limit dive performance (duration and depth) in king penguins (Aptenodytes patagonicus). The model allowed prediction of blood and tissue (central circulation, muscle, brain and fat) N2 tensions (P(N2)) based on different cardiac outputs and blood flow distributions. Estimated mixed venous P(N2) agreed with values observed during forced dives in a compression chamber used to validate the assumptions of the model. During bouts of foraging dives, estimated mixed venous and tissue P(N2) increased as the bout progressed. Estimated mean maximum mixed venous P(N2) upon return to the surface after a dive was 4.56+/-0.18 atmospheres absolute (ATA; range: 4.37-4.78 ATA). This is equivalent to N2 levels causing a 50% DCS incidence in terrestrial animals of similar mass. Bout termination events were not associated with extreme mixed venous N2 levels. Fat P(N2) was positively correlated with bout duration and the highest estimated fat P(N2) occurred at the end of a dive bout. The model suggested that short and shallow dives occurring between dive bouts help to reduce supersaturation and thereby DCS risk. Furthermore, adipose tissue could also help reduce DCS risk during the first few dives in a bout by functioning as a sink to buffer extreme levels of N2.

Metabolism and thermoregulation during fasting in king penguins,Aptenodytes patagonicus,in air and water

We measured oxygen consumption rate (V̇o2) and body temperatures in 10 king penguins in air and water. V̇o2was measured during rest and at submaximal and maximal exercise before (fed) and after (fasted) an average fasting duration of 14.4 ± 2.3 days (mean ± 1 SD, range 10–19 days) in air and water. Concurrently, we measured subcutaneous temperature and temperature of the upper (heart and liver), middle (stomach) and lower (intestine) abdomen. The mean body mass (Mb) was 13.8 ± 1.2 kg in fed and 11.0 ± 0.6 kg in fasted birds. After fasting, resting V̇o2was 93% higher in water than in air (air: 86.9 ± 8.8 ml/min; water: 167.3 ± 36.7 ml/min, P < 0.01), while there was no difference in resting V̇o2between air and water in fed animals (air: 117.1 ± 20.0 ml O2/min; water: 114.8 ± 32.7 ml O2/min, P > 0.6). In air, V̇o2decreased with Mb, while it increased with Mbin water. Body temperature did not change with fasting in air, whereas in water, there were complex changes in the peripheral body temperatures. These latter changes may, therefore, be indicative of a loss in body insulation and of variations in peripheral perfusion. Four animals were given a single meal after fasting and the temperature changes were partly reversed 24 h after refeeding in all body regions except the subcutaneous, indicating a rapid reversal to a prefasting state where body heat loss is minimal. The data emphasize the importance in considering nutritional status when studying king penguins and that the fasting-related physiological changes diverge in air and water.

Heart rate and energetics of free-ranging king penguins (Aptenodytes patagonicus)

The main objective of this study was to determine heart rate (fh) and the energetic costs of specific behaviours of king penguins while ashore and while foraging at sea during their breeding period. In particular, an estimate was made of the energetic cost of diving in order to determine the proportion of dives that may exceed the calculated aerobic dive limit (cADL; estimated usable O2 stores/estimated rate of oxygen consumption during diving). An implanted data logger enabled fh and diving behaviour to be monitored from 10 free-ranging king penguins during their breeding period. Using previously determined calibration equations, it was possible to estimate rate of oxygen consumption (VO2) when the birds were ashore and during various phases of their foraging trips. Diving behaviour showed a clear diurnal pattern, with a mixture of deep (>40 m), long (>3 min) and shallow (<40 m), short (<3 min) dives from dawn to dusk and shallow, short dives at night. Heart rate during dive bouts and dive cycles (dive + post-dive interval) was 42% greater than that when the birds were ashore. During diving, fh was similar to the 'ashore' value (87+/-4 beats min(-1)), but it did decline to 76% of the value recorded from king penguins resting in water. During the first hour after a diving bout, fh was significantly higher than the average value during diving (101+/-4 beats min(-1)) and for the remainder of the dive bout. Rates of oxygen consumption estimated from these (and other) values of fh indicate that when at sea, metabolic rate (MR) was 83% greater than that when the birds were ashore [3.15 W kg(-1) (-0.71, +0.93), where the values in parentheses are the computed standard errors of the estimate], while during diving bouts and dive cycles, it was 73% greater than the 'ashore' value. Although estimated MR during the total period between dive bouts was not significantly different from that during dive bouts [5.44 W kg(-1) (-0.30, +0.32)], MR during the first hour following a dive bout was 52% greater than that during a diving bout. It is suggested that this large increase following diving (foraging) activity is, at least in part, the result of rewarming the body, which occurs at the end of a diving bout. From the measured behaviour and estimated values of VO2, it was evident that approximately 35% of the dives were in excess of the cADL. Even if VO2 during diving was assumed to be the same as when the birds were resting on water, approximately 20% of dives would exceed the cADL. As VO2 during diving is, in fact, that estimated for a complete dive cycle, it is quite feasible that VO2 during diving itself is less than that measured for birds resting in water. It is suggested that the regional hypothermia that has been recorded in this species during diving bouts may be at least a contributing factor to such hypometabolism.

Effect of fasting on the V̇o2-fhrelationship in king penguins,Aptenodytes patagonicus

King penguins ( Aptenodytes patagonicus) may fast for up to 30 days during their breeding period. As such extended fasting may affect the relationship between the rate of O2consumption (V̇o2) and heart rate ( fH), five male king penguins were exercised at various speeds on repeated occasions during a fasting period of 24–31 days. In addition, V̇o2and fHwere measured in the same animals during rest in cold air and water (4°C). V̇o2and fHat rest and V̇o2during exercise decreased with fasting. There was a significant relation between V̇o2and fH( r2= 0.56) that was improved by including speed, body mass ( Mb), number of days fasting ( t), and a cross term between fHand t ( r2= 0.92). It was concluded that there was a significant change in the V̇o2- fHrelationship with fasting during exercise. As t is measurable in the field and was shown to be significant and, therefore, a practical covariate, a regression equation for use when birds are ashore was obtained by removing speed and Mb. When this equation was used, predicted V̇o2was in good agreement with the observed data, with an overall error of 3.0%. There was no change in the V̇o2- fHrelationship in penguins at rest in water.

The heart rate/oxygen consumption relationship during cold exposure of the king penguin: a comparison with that during exercise

This study investigated whether exposure to low ambient temperature could be used as an alternative to exercise for calibrating heart rate (fH)against rate of oxygen consumption(V̇O2) for subsequent use of fH to estimate V̇O2 in free-ranging animals. Using the relationship between the oxygen pulse (OP, the amount of oxygen used per heart beat) and an index of body condition (or nutritional index, NI), a relationship between fH and V̇O2 was established for resting king penguins exposed to a variety of environmental temperatures. Although there was a small but significant increase in the OP above and below the lower critical temperature (-4.9°C), there was no difference in the relationship obtained between the OP and body condition (NI)obtained above or below the lower critical temperature. These results were then compared with those obtained in a previous study in which the relationship between fH and V̇O2 had been established for king penguins during steady-state exercise. The relationship between OP and NI in the present study was not significantly different from the relationship between resting OP and NI in the previous study. However, the relationship was different from that between active OP and NI. We conclude that, at least for king penguins, although thermoregulation does not affect the relationship between resting OP and NI, temperature cannot be used as an alternative to exercise for calibrating fH against V̇O2 for subsequent use of fH to estimate V̇O2 in free-ranging animals.

Buoyancy and maximal diving depth in penguins

Using a newly developed data logger to measure acceleration, we demonstrate that free-ranging king and Adélie penguins only beat their flippers substantially during the first part of descent or when they were presumed to be chasing prey at the bottom of dives. Flipper beating stopped during the latter part of ascent: at 29±9 % (mean ± S.D.) of dive depth(mean dive depth=136.8±145.1 m, N=425 dives) in king penguins,and at 52±20 % of dive depth (mean dive depth=72.9±70.5 m, N=664 dives) in Adélie penguins. Propulsive swim speeds of both species were approximately 2 m s-1 during dives; however, a marked increase in speed, up to approximately 2.9 m s-1, sometimes occurred in king penguins during the passive ascending periods. During the prolonged ascending, oblique ascent angle and slowdown near the surface may represent one way to avoid the potential risk of decompression sickness. Biomechanical calculations for data from free-ranging king and Adélie penguins indicate that the air volume of the birds (respiratory system and plumage) can provide enough buoyancy for the passive ascent. When comparing the passive ascents for shallow and deep dives, there is a positive correlation between air volume and the depth of the dive. This suggests that penguins regulate their air volume to optimize the costs and benefits of buoyancy.

Heart rate as an indicator of oxygen consumption: influence of body condition in the king penguin

The use of heart rate to estimate field metabolic rate has become a more widely used technique. However, this method also has some limitations, among which is the possible impact that several variables such as sex, body condition (i.e. body fat stores) and/or inactivity might have on the relationship between heart rate and rate of oxygen consumption. In the present study, we investigate the extent to which body condition can affect the use of heart rate as an indicator of the rate of oxygen consumption.

Twenty-two breeding king penguins (Aptenodytes patagonicus) were exercised on a variable-speed treadmill. These birds were allocated to four groups according to their sex and whether or not they had been fasting. Linear regression equations were used to describe the relationship between heart rate and the rate of oxygen consumption for each group. There were significant differences between the regression equations for the four groups.

Good relationships were obtained between resting and active oxygen pulses and an index of the body condition of the birds. Validation experiments on six courting king penguins showed that the use of a combination of resting oxygen pulse and active oxygen pulse gave the best estimate of the rate of oxygen consumption V̇O2. The mean percentage error between predicted and measured V̇O2 was only +0.81% for the six birds.

We conclude that heart rate can be used to estimate rate of oxygen consumption in free-ranging king penguins even over a small time scale (30min). However, (i) the type of activity of the bird must be known and (ii) the body condition of the bird must be accurately determined. More investigations on the impact of fasting and/or inactivity on this relationship are required to refine these estimates further.

Feeding behaviour of free-ranging penguins determined by oesophageal temperature

Sea birds play a major role in marine food webs, and it is important to determine when and how much they feed at sea. A major advance has been made by using the drop in stomach temperature after ingestion of ectothermic prey. This method is less sensitive when birds eat small prey or when the stomach is full. Moreover, in diving birds, independently of food ingestion, there are fluctuations in the lower abdominal temperature during the dives. Using oesophageal temperature, we present here a new method for detecting the timing of prey ingestion in free-ranging sea birds, and, to our knowledge, report the first data obtained on king penguins (Aptenodytes patagonicus). In birds ashore, which were hand-fed 2-15 g pieces of fish, all meal ingestions were detected with a sensor in the upper oesophagus. Detection was poorer with sensors at increasing distances from the beak. At sea, slow temperature drops in the upper oesophagus and stomach characterized a diving effect per se. For the upper oesophagus only, abrupt temperature variations were superimposed, therefore indicating prey ingestions. We determined the depths at which these occurred. Combining the changes in oesophageal temperatures of marine predators with their diving pattern opens new perspectives for understanding their foraging strategy, and, after validation with concurrent applications of classical techniques of prey survey, for assessing the distribution of their prey.

Impact of externally attached loggers on the diving behaviour of the king penguin

The impact of relatively small externally attached time series recorders on some foraging parameters of seabirds was investigated during the austral summer of 1995 by monitoring the diving behaviour of 10 free-ranging king penguins (Aptenodytes patagonicus) over one foraging trip. Time-depth recorders were implanted in the abdominal cavities of the birds, and half of the animals also had dummy loggers attached on their backs. Although most of the diving behaviour was not significantly affected by the external loggers (P>0.05), the birds with externally attached loggers performed almost twice as many shallow dives, between 0 and 10 m depth, as the birds without external loggers. These shallow dives interrupted more frequently the deep-diving sequences in the case of birds with external loggers (percentage of deep dives followed by deep dives: 46% for birds with implants only vs. 26% for birds with an external attachment). Finally, the distribution pattern of the postdive durations plotted against the hour of the day was more heterogeneous for the birds with an external package. In addition, these penguins had extended surfacing times between two deep dives compared to birds without external attachments (P<0.0001). These results suggest the existence of an extra energy cost induced by externally attached loggers.

Body reserves and nutritional needs during laying preparation in barn owls

To investigate the composition of the body reserves made during pre-laying by breeding European barn owls (Tyto alba), we have analysed the body composition of captive breeding and non-breeding females sacrificed during the laying period. The data obtained were compared to the daily requirement for egg formation obtained by an egg composition analysis and the timing of egg synthesis. This study demonstrates that body mass gain observed in breeding females (+ 38.3 g after eggs in formation and gonadal tractus were removed) was not the consequence of an accumulation of body fuels like lipids but of mainly water and lean material. The lipidic reserves were found to be less important in breeding than in non-breeding females and their localisation modified; lipids were absent from medullar bones in breeding females which liberated room for other storage. The subcutaneous tissue, which was homogeneous in non-breeding females, was located principally under the brood patch in breeding females. Nutrients and energy required during egg formation could be obtained without modification of daily food intake. These results show that a laying event can be initiated in 14 days and that the onset of reproduction is not triggered by body condition in barn owls. The water accumulation observed is suggested to be the mere consequence of an increase of protein metabolism (egg and moult). These results intimate that the body mass increase observed in diurnal and nocturnal raptors during laying preparation, interpreted as an energy safety strategy, ought to be reconsidered.

Energy metabolism and body temperature of barn owls fasting in the cold

Energetic adaptation to fasting in the cold has been investigated in a nocturnal raptor, the barn owl (Tyto alba), during winter. Metabolic rate and body temperature (Tb) were monitored in captive birds, (1) after acute exposure to different ambient temperatures (Ta), and (2) during a prolonged fast in the cold (4 degrees C), to take into account the three characteristic phases of body fuel utilization that occur during a long-term but reversible fast. In postabsorptive birds, metabolic rate in the thermoneutral zone was 4. 1+/-0.1 W kg-1 and increased linearly below a lower critical temperature of 23 degrees C. Metabolic rate was 70% above basal at +4 degrees C Ta. Wet thermal conductance was 0.22 W kg-1 degrees C-1. During fasting in the cold, the mass-specific resting metabolic rate decreased by 16% during the first day (phase I) and remained constant thereafter. The amplitude of the daily rhythm in Tb was only moderately increased during phase II, with a slight lowering (0. 6 degrees C) in minimal diurnal Tb, but rose markedly in phase III with a larger drop (1.4 degrees C) in minimal diurnal Tb. Refeeding the birds ended phase III and reversed the observed changes. These results indicate that diurnal hypothermia may be used in long-term fasting barn owls and could be triggered by a threshold of body lipid depletion, according to the shift from lipid to protein fuel metabolism occurring at the phase II/phase III transition. The high cost of regulatory thermogenesis and the limited use of hypothermia during fasting may contribute to the high mortality of barn owls during winter.

Stress in birds due to routine handling and a technique to avoid it

The stress that might result in animals from the routine handling that most experimental studies involve, e.g., weighing, injecting, and blood sampling, is usually assumed to be minimal when the animals look quiet. However, the intensity of this stress remains largely ignored. We have developed a system that allows blood samples to be taken from freely behaving geese without entering the animal room. In these entirely undisturbed geese, the humoral indexes of stress, i.e., blood levels of catecholamines, corticosterone, and lactate, were as low or even lower than the lowest values previously reported for birds. Remarkably, the mean basal values for epinephrine and norepinephrine were 90-fold and 5-fold, respectively, below the lowest values in the literature. Stress-induced variations in pH that would have concealed detection of nutrition-induced changes in pH were eliminated. In contrast, even though the birds looked quiet during a short 5-min routine handling procedure, to which they had been accustomed for weeks, there was a dramatic increase in the level of humoral indexes of stress. These increased severalfold within only 2 min, and the return to initial values could take up to 1 h. Acid-base balance was also disrupted. Thus, in studies on animals, the absence of stress cannot be deduced from only behavioral observations. Only a system for taking blood without human interference may enable stress-free investigations.

Gas diffusive conductance of sea-level hen eggs incubated at 2900 m altitude

Sea level hen eggs, selected for their shell conductance (water vapor conductance, HH2O), were incubated at a simulated high altitude, PB = 529 Torr, ca. 2900 m, at 72% relative humidity (rh) to prevent excessive water loss due to hypobaric condition; they were transferred to 150 m 18-24 h before measurements. Control eggs were incubated at 150 m, PB = 750 Torr, rh = 60%. In 8- to 18-day embryos, total CO diffusive conductance, GCO; embyronic body mass, BM; oxygen consumption, MO2; blood hematocrit, Hct; hemoglobin concentration, [Hb]; and heart mass, HM, were measured. Total water loss was the same in both groups, 12% initial egg mass. However, the severe effects of high altitude: 72% mortality and 9% malformation, and reduced increases of BM and MO2, can be related partially to the strong hypocapnia, which resulted from the high shell conductance (GH2O = 18.1 mg X (d X Torr)-1, and was superimposed on the hypoxia. GCO was reduced, while Hct, [Hb] and HM were not significantly affected. When measurements were normalized to BM, MO2 and GCO were identical in the two groups, whereas [Hb] and HM were higher at 2900 m (differential growth). Thus, during incubation, gas diffusive conductance appeared to depend on embryo development and did not adapt to altitude hypoxia. Compared with controls, GCO in high-altitude eggs actually decreased in proportion to BM growth.