Behavioural inference from signal processing using animal-borne multi-sensor loggers: a novel solution to extend the knowledge of sea turtle ecology

The identification of sea turtle behaviours is a prerequisite to predicting the activities and time-budget of these animals in their natural habitat over the long term. However, this is hampered by a lack of reliable methods that enable the detection and monitoring of certain key behaviours such as feeding. This study proposes a combined approach that automatically identifies the different behaviours of free-ranging sea turtles through the use of animal-borne multi-sensor recorders (accelerometer, gyroscope and time-depth recorder), validated by animal-borne video-recorder data. We show here that the combination of supervised learning algorithms and multi-signal analysis tools can provide accurate inferences of the behaviours expressed, including feeding and scratching behaviours that are of crucial ecological interest for sea turtles. Our procedure uses multi-sensor miniaturized loggers that can be deployed on free-ranging animals with minimal disturbance. It provides an easily adaptable and replicable approach for the long-term automatic identification of the different activities and determination of time-budgets in sea turtles. This approach should also be applicable to a broad range of other species and could significantly contribute to the conservation of endangered species by providing detailed knowledge of key animal activities such as feeding, travelling and resting.

Heart rate and cardiac response to exercise during voluntary dives in captive sea turtles (Cheloniidae)

In chelonids, oxygen is primarily stored in the lungs during a dive. Therefore, management of blood oxygen transportation to peripheral tissues by cardiovascular adjustments during submergence is crucial to maximize their dive duration, and consequently, the time spent for ecological activities such as foraging. However, the cardiac response to exercise has rarely been examined in sea turtles. In this study, heart rate and its relationship with exercise during voluntary dives were determined in six captive green turtles (19.4±1.5 kg) by simultaneously recording depth, acceleration and electrocardiogram. Our results demonstrated that the heart rate of green turtles was generally low (11.1±0.4 bpm) during resting dives, but they often exhibited instantaneously extreme tachycardia (up to 78.4 bpm). Green turtles elevated their heart rate up to 39.8±1.5 bpm during ventilation after resting dives, while up to 33.1±1.4 bpm after active dives. The heart rate immediately elevated with onset of exercise, and increased linearly with exercise. This result may indicate that turtles immediately need to transport oxygen from the lungs to peripheral tissues by pulmonary and systemic circulations to meet the metabolic demands of exercise because they mainly store oxygen in their lungs.

Assessing the effect of recreational scallop harvest on the distribution and behaviour of foraging marine turtles

The impact of fisheries on marine megafauna is widely known but most studies have focused on commercial fisheries, overlooking the effect of local recreational fisheries. This is particularly important for marine turtles in near-shore habitats that overlap with recreational fisheries. We assessed the effect of recreational scallop fisheries on the distribution and behaviour of foraging marine turtles in the coastal waters of the upper Eastern Gulf of Mexico. Before and during the scallop season we quantified the density and overlap of marine turtles and vessels sighted, and satellite tracked four turtles to assess their distribution and behaviour. The relative distribution of marine turtles sighted during the scallop season overlapped with 48% of the area most frequently used by harvesters, and marine turtle activity hotspots shifted between seasons. In addition, during the scallop season the home range size of individual turtles appeared to decrease, and turtles displayed frequent changes in travel speed and directionality. We hypothesize that such changes are probably related to the distribution and movement of vessels and the abundant presence of people in the water. Our study highlights the importance of considering recreational fisheries and their local effect on marine megafauna for informing future adaptive management practices. However, further studies are needed to quantify the direct and indirect impacts of recreational fisheries and to assess the degree of risk of associated activities to marine turtle populations.

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.

Development and application of a machine learning algorithm for classification of elasmobranch behaviour from accelerometry data

Discerning behaviours of free-ranging animals allows for quantification of their activity budget, providing important insight into ecology. Over recent years, accelerometers have been used to unveil the cryptic lives of animals. The increased ability of accelerometers to store large quantities of high resolution data has prompted a need for automated behavioural classification. We assessed the performance of several machine learning (ML) classifiers to discern five behaviours performed by accelerometer-equipped juvenile lemon sharks (Negaprion brevirostris) at Bimini, Bahamas (25°44'N, 79°16'W). The sharks were observed to exhibit chafing, burst swimming, headshaking, resting and swimming in a semi-captive environment and these observations were used to ground-truth data for ML training and testing. ML methods included logistic regression, an artificial neural network, two random forest models, a gradient boosting model and a voting ensemble (VE) model, which combined the predictions of all other (base) models to improve classifier performance. The macro-averaged F-measure, an indicator of classifier performance, showed that the VE model improved overall classification (F-measure 0.88) above the strongest base learner model, gradient boosting (0.86). To test whether the VE model provided biologically meaningful results when applied to accelerometer data obtained from wild sharks, we investigated headshaking behaviour, as a proxy for prey capture, in relation to the variables: time of day, tidal phase and season. All variables were significant in predicting prey capture, with predations most likely to occur during early evening and less frequently during the dry season and high tides. These findings support previous hypotheses from sporadic visual observations.

Surfacers change their dive tactics depending on the aim of the dive: evidence from simultaneous measurements of breaths and energy expenditure

Air-breathing divers are assumed to have evolved to apportion their time between surface and underwater periods to maximize the benefit gained from diving activities. However, whether they change their time allocation depending on the aim of the dive is still unknown. This may be particularly crucial for 'surfacers' because they dive for various purposes in addition to foraging. In this study, we counted breath events at the surface and estimated oxygen consumption during resting, foraging and other dives in 11 green turtles (Chelonia mydas) in the wild. Breath events were counted by a head-mounted acceleration logger or direct observation based on an animal-borne video logger, and oxygen consumption was estimated by measuring overall dynamic body acceleration. Our results indicate that green turtles maximized their submerged time, following this with five to seven breaths to replenish oxygen for resting dives. However, they changed their dive tactic during foraging and other dives; they surfaced without depleting their estimated stores of oxygen, followed by only a few breaths for effective foraging and locomotion. These dichotomous surfacing tactics would be the result of behavioural modifications by turtles depending on the aim of each dive.

Climate and the individual: inter-annual variation in the autumnal activity of the European badger (Meles meles)

We establish intra-individual and inter-annual variability in European badger (Meles meles) autumnal nightly activity in relation to fine-scale climatic variables, using tri-axial accelerometry. This contributes further to understanding of causality in the established interaction between weather conditions and population dynamics in this species. Modelling found that measures of daylight, rain/humidity, and soil temperature were the most supported predictors of ACTIVITY, in both years studied. In 2010, the drier year, the most supported model included the SOLAR*RH interaction, RAIN, and30cmTEMP (w = 0.557), while in 2012, a wetter year, the most supported model included the SOLAR*RH interaction, and the RAIN*10cmTEMP (w = 0.999). ACTIVITY also differed significantly between individuals. In the 2012 autumn study period, badgers with the longest per noctem activity subsequently exhibited higher Body Condition Indices (BCI) when recaptured. In contrast, under drier 2010 conditions, badgers in good BCI engaged in less per noctem activity, while badgers with poor BCI were the most active. When compared on the same calendar dates, to control for night length, duration of mean badger nightly activity was longer (9.5 hrs ±3.3 SE) in 2010 than in 2012 (8.3 hrs ±1.9 SE). In the wetter year, increasing nightly activity was associated with net-positive energetic gains (from BCI), likely due to better foraging conditions. In a drier year, with greater potential for net-negative energy returns, individual nutritional state proved crucial in modifying activity regimes; thus we emphasise how a ‘one size fits all’ approach should not be applied to ecological responses.

Ethogram of Immature Green Turtles: Behavioral Strategies for Somatic Growth in Large Marine Herbivores

Animals are assumed to obtain/conserve energy effectively to maximise their fitness, which manifests itself in a variety of behavioral strategies. For marine animals, however, these behavioral strategies are generally unknown due to the lack of high-resolution monitoring techniques in marine habitats. As large marine herbivores, immature green turtles do not need to allocate energy to reproduction but are at risk of shark predation, although it is a rare occurrence. They are therefore assumed to select/use feeding and resting sites that maximise their fitness in terms of somatic growth, while avoiding predation. We investigated fine-scale behavioral patterns (feeding, resting and other behaviors), microhabitat use and time spent on each behavior for eight immature green turtles using data loggers including: depth, global positioning system, head acceleration, speed and video sensors. Immature green turtles at Iriomote Island, Japan, spent an average of 4.8 h feeding on seagrass each day, with two peaks, between 5∶00 and 9∶00, and between 17∶00 and 20∶00. This feeding pattern appeared to be restricted by gut capacity, and thus maximised energy acquisition. Meanwhile, most of the remaining time was spent resting at locations close to feeding grounds, which allowed turtles to conserve energy spent travelling and reduced the duration of periods exposed to predation. These behavioral patterns and time allocations allow immature green turtles to effectively obtain/conserve energy for growth, thus maximising their fitness.

Loggerhead turtles (Caretta caretta) use vision to forage on gelatinous prey in mid-water

Identifying characteristics of foraging activity is fundamental to understanding an animals’ lifestyle and foraging ecology. Despite its importance, monitoring the foraging activities of marine animals is difficult because direct observation is rarely possible. In this study, we use an animal-borne imaging system and three-dimensional data logger simultaneously to observe the foraging behaviour of large juvenile and adult sized loggerhead turtles (Caretta caretta) in their natural environment. Video recordings showed that the turtles foraged on gelatinous prey while swimming in mid-water (i.e., defined as epipelagic water column deeper than 1 m in this study). By linking video and 3D data, we found that mid-water foraging events share the common feature of a marked deceleration phase associated with the capture and handling of the sluggish prey. Analysis of high-resolution 3D movements during mid-water foraging events, including presumptive events extracted from 3D data using deceleration in swim speed as a proxy for foraging (detection rate = 0.67), showed that turtles swam straight toward prey in 171 events (i.e., turning point absent) but made a single turn toward the prey an average of 5.7±6.0 m before reaching the prey in 229 events (i.e., turning point present). Foraging events with a turning point tended to occur during the daytime, suggesting that turtles primarily used visual cues to locate prey. In addition, an incident of a turtle encountering a plastic bag while swimming in mid-water was recorded. The fact that the turtle’s movements while approaching the plastic bag were analogous to those of a true foraging event, having a turning point and deceleration phase, also support the use of vision in mid-water foraging. Our study shows that integrated video and high-resolution 3D data analysis provides unique opportunities to understand foraging behaviours in the context of the sensory ecology involved in prey location.

From sensor data to animal behaviour: an oystercatcher example

Animal-borne sensors enable researchers to remotely track animals, their physiological state and body movements. Accelerometers, for example, have been used in several studies to measure body movement, posture, and energy expenditure, although predominantly in marine animals. In many studies, behaviour is often inferred from expert interpretation of sensor data and not validated with direct observations of the animal. The aim of this study was to derive models that could be used to classify oystercatcher (Haematopus ostralegus) behaviour based on sensor data. We measured the location, speed, and tri-axial acceleration of three oystercatchers using a flexible GPS tracking system and conducted simultaneous visual observations of the behaviour of these birds in their natural environment. We then used these data to develop three supervised classification trees of behaviour and finally applied one of the models to calculate time-activity budgets. The model based on accelerometer data developed to classify three behaviours (fly, terrestrial locomotion, and no movement) was much more accurate (cross-validation error = 0.14) than the model based on GPS-speed alone (cross-validation error = 0.35). The most parsimonious acceleration model designed to classify eight behaviours could distinguish five: fly, forage, body care, stand, and sit (cross-validation error = 0.28); other behaviours that were observed, such as aggression or handling of prey, could not be distinguished. Model limitations and potential improvements are discussed. The workflow design presented in this study can facilitate model development, be adapted to a wide range of species, and together with the appropriate measurements, can foster the study of behaviour and habitat use of free living animals throughout their annual routine.

Detection of feeding behaviour in common carp Cyprinus carpio by using an acceleration data logger to identify mandibular movement

Miniaturized acceleration data loggers were attached to the lower mandible of common carp Cyprinus carpio to remotely identify feeding behaviour. Whether the acceleration signal could distinguish the quantity and quality of food was also investigated. The frequency and amplitude of the lower mandible stroke, calculated from surging acceleration determined by continuous wavelet transformation, significantly increased during the feeding period compared to that during the non-feeding period. These characteristic movement patterns were maintained for mean ±s.e. 187·3 ± 38·2 s when the fish were fed a single item of food and for mean ±s.e. 419·3 ± 28·6 s when they consumed multiple items. The dominant cycle and amplitude calculated according to feeding event duration, however, did not differ significantly between the two types of diets the fish consumed. Surging acceleration could detect mean ±s.e. 89·8 ± 13·5% of feeding events, although the false detection rate was mean ±s.e. 25·9 ± 10·9%. The results indicate that the mandible acceleration measurement method could be utilized to detect and record the feeding events in fishes that use a suction feeding mode similar to C. carpio.

Identification of distinct movement patterns in Pacific leatherback turtle populations influenced by ocean conditions

Interactions with fisheries are believed to be a major cause of mortality for adult leatherback turtles (Dermochelys coriacea), which is of particular concern in the Pacific Ocean, where they have been rapidly declining. In order to identify where these interactions are occurring and how they may be reduced, it is essential first to understand the movements and behavior of leatherback turtles. There are two regional nesting populations in the East Pacific (EP) and West Pacific (WP), comprising multiple nesting sites. We synthesized tracking data from the two populations and compared their movement patterns. A switching state-space model was applied to 135 Argos satellite tracks to account for observation error, and to distinguish between migratory and area-restricted search behaviors. The tracking data, from the largest leatherback data set ever assembled, indicated that there was a high degree of spatial segregation between EP and WP leatherbacks. Area-restricted search behavior mainly occurred in the southeast Pacific for the EP leatherbacks, whereas the WP leatherbacks had several different search areas in the California Current, central North Pacific, South China Sea, off eastern Indonesia, and off southeastern Australia. We also extracted remotely sensed oceanographic data and applied a generalized linear mixed model to determine if leatherbacks exhibited different behavior in relation to environmental variables. For the WP population, the probability of area-restricted search behavior was positively correlated with chlorophyll-a concentration. This response was less strong in the EP population, but these turtles had a higher probability of search behavior where there was greater Ekman upwelling, which may increase the transport of nutrients and consequently prey availability. These divergent responses to oceanographic conditions have implications for leatherback vulnerability to fisheries interactions and to the effects of climate change. The occurrence of leatherback turtles within both coastal and pelagic areas means they have a high risk of exposure to many different fisheries, which may be very distant from their nesting sites. The EP leatherbacks have more limited foraging grounds than the WP leatherbacks, which could make them more susceptible to any temperature or prey changes that occur in response to climate change.