Monitoring Animal Behavior Research Articles

Tracking group housed sows with an ultra-wideband indoor positioning system: A feasibility study

Automated tracking of indoor housed farm animals is gaining an increasing interest in farming practice and research for monitoring animal behaviour and health. Several positioning solutions have become commercially available for different animal species, but the insight on how these systems perform in pig facilities is still limited. In this study the feasibility of tracking sows in a barn using an ultra-wideband (UWB) based indoor positioning system was investigated. The system consisted of 7 base stations and multiple tracker tags, and the determination of tags' positions was based on time of arrival (ToA). For locating stationary tags the system achieved 0.37 m accuracy at 1 m height and 0.50 m accuracy at 0.3 m height. While tracking moving tags at 1 m height 0.38 m accuracy was achieved. Three filters were tested to further improve the positioning performance. The median filter had the highest improvement on locating stationary and moving tags. The bilateral filter had a balanced capability of uncovering complex moving trajectories. The double exponential moving average filter was more suited when real-time updates were required. The overall results showed that tracking group housed sows inside a pig barn was feasible using the UWB positioning system. • A UWB positioning system achieved <0.5 m positioning accuracy using in a sow barn. • The positioning accuracy was dependent on the location and height of tracker tags. • Filters were more effective for improving tracking precision than accuracy. • The choice of filters depends on the motion characteristics and the application.

An open-source research tool to study triaxial inertial sensors for monitoring selected behaviors in sheep.

The use of automated systems for monitoring animal behavior provides information on individual animal behavior and can be used to enhance animal productivity. However, the advancement of this industry is hampered by technology costs, challenges with power supplies, limited data accessibility, and inconsistent testing approaches for confirming the detection of livestock behaviors. Development of open-source research tools similar to commercially available wearable technologies may contribute to the development of more-efficient and affordable technologies. The objective of this study was to demonstrate an open-source, microprocessor-based sensor designed to monitor and enable differentiation among selected behaviors of adult wethers. The sensor was comprised of an inexpensive espressif ESP-32-WROOM-32 microprocessor with Bluetooth communication, a generic MPU92/50 motion sensor that contains a three-axis accelerometer, three-axis magnetometer, a three-axis gyroscope, and a 5-V rechargeable lithium-ion battery. The open-source Arduino IDE software was used to program the microprocessor and to adjust the frequency of sampling, the data packet to send, and the operating conditions. For demonstration purposes, sensors were placed on six housed sheep for three 1-h increments with concurrent visual behavioral observation. Sensor readings (x-, y-, and z-axis) were summarized (mean and SD) within a minute and compared to animal behavior observations (also on a by-minute basis) using a linear mixed-effect model with animal as a random effect and behavioral classifier as a fixed effect. This analysis demonstrated the basic utility of the sensor to differentiate among animal behaviors based on sensed data (P < 0.001). Although substantial additional work is needed for algorithm development, power source testing, and network optimization, this open-source platform appears to be a promising strategy to research wearable sensors in a generalizable manner.

Quantifying the Dynamics of Livestock Distribution by Unmanned Aerial Vehicles (UAVs): A Case Study of Yak Grazing at the Household Scale

Managed grazing is the most extensive land use worldwide. Information about the spatio-temporal distribution of livestock is critical for grassland ecosystem management. However, no direct and cost-effective method exists to monitor livestock distribution under natural conditions. We proposed a practical method that uses unmanned aerial vehicles (UAVs), and tested it at a typical household pasture on the Qinghai-Tibetan Plateau. We monitored and analyzed the hourly spatial distribution of each yak (Bos mutus). We found that (1) yak information extracted by our independently developed software was sufficiently accurate (P > 0.05); (2) yak density (frequency/area/period) decreased with increasing distance from campsites in the early growing season, while the pattern reversed in the late growing season; (3) dispersion index (DI, m2/yak) of the yak herd exhibited a tendency to decrease in the morning and then to increase in the afternoon within a day; from June to September DI decreased in both the morning and afternoon, but was similar at noon. The monitoring method we developed in this study characterized the spatial distribution of the whole yak herd dynamically with high frequency, resolution, and efficiency. We conclude that this method is suitable for long-term monitoring of animal behavior, and for studying relationship between the distribution of pastoral livestock and resource availability, which will be beneficial for grassland sustainable management.

A sensor-fusion-system for tracking sheep location and behaviour

The growing interest in precision livestock farming is prompted by a desire to understand the basic behavioural needs of the animals and optimize the contribution of each animal. The aim of this study was to develop a system that automatically generated individual animal behaviour and localization data in sheep. A sensor-fusion-system tracking individual sheep position and detecting sheep standing/lying behaviour was proposed. The mean error and standard deviation of sheep position performed by the ultra-wideband location system was 0.357 ± 0.254 m, and the sensitivity of the sheep standing and lying detection performed by infrared radiation cameras and three-dimenional computer vision technology were 98.16% and 100%, respectively. The proposed system was able to generate individual animal activity reports and the real-time detection was achieved. The system can increase the convenience for animal behaviour studies and monitoring of animal welfare in the production environment.

Repeatability analysis improves the reliability of behavioral data.

Reliability of data has become a major concern in the course of the reproducibility crisis. Especially when studying animal behavior, confounding factors such as novelty of the test apparatus can lead to a wide variability of data which may mask treatment effects and consequently lead to misinterpretation. Habituation to the test situation is a common practice to circumvent novelty induced increases in variance and to improve the reliability of the respective measurements. However, there is a lack of published empirical knowledge regarding reasonable habituation procedures and a method validation seems to be overdue. This study aimed at setting up a simple strategy to increase reliability of behavioral data measured in a familiar test apparatus. Therefore, exemplary data from mice tested in an Open Field (OF) arena were used to elucidate the potential of habituation and how reliability of measures can be confirmed by means of a repeatability analysis using the software R. On seven consecutive days, male C57BL/6J, BALB/cJ and 129S1/SvImJ mice were tested in an OF arena once daily and individual mouse behavior was recorded. A repeatability analysis was conducted with regard to repeated trials of habituation. Our data analysis revealed that monitoring animal behavior during habituation is important to determine when individual differences of the measurements are stable. Repeatability values from distance travelled and average activity increased over the habituation period, revealing that around 60% of the variance of the data can be explained by individual differences between mice. The first day of habituation was significantly different from the following 6 days. A three-day habituation period appeared to be sufficient in this study. Overall, these results emphasize the importance of habituation and in depth analysis of habituation data to define the correct starting point of the experiment for improving the reliability and reproducibility of experimental data.

Thinking small: Next-generation sensor networks close the size gap in vertebrate biologging.

Recent advances in animal tracking technology have ushered in a new era in biologging. However, the considerable size of many sophisticated biologging devices restricts their application to larger animals, whereas older techniques often still represent the state-of-the-art for studying small vertebrates. In industrial applications, low-power wireless sensor networks (WSNs) fulfill requirements similar to those needed to monitor animal behavior at high resolution and at low tag mass. We developed a wireless biologging network (WBN), which enables simultaneous direct proximity sensing, high-resolution tracking, and long-range remote data download at tag masses of 1 to 2 g. Deployments to study wild bats created social networks and flight trajectories of unprecedented quality. Our developments highlight the vast capabilities of WBNs and their potential to close an important gap in biologging: fully automated tracking and proximity sensing of small animals, even in closed habitats, at high spatial and temporal resolution.

Machine learning to detect behavioural anomalies in dairy cows under subacute ruminal acidosis

Sickness behaviour is characterised by a lethargic state during which the animal reduces its activity, sleeps more and at times when normally awake, reduces its feed and water intake, and interacts less with its environment. Subtle modifications in behaviour can materialise just before clinical signs of a disease. Recent sensor developments enable continuous monitoring of animal behaviour, but the shift to abnormal animal activity remains difficult to detect. We explored the use of Machine Learning (ML) to detect abnormal behaviour from continuous monitoring. We submitted 14 cows (Bos taurus) to Sub-Acute Ruminal Acidosis (SARA), a disease known to induce changes in behaviour. Another 14 control cows were not submitted to SARA. We used a ruminal bolus to monitor pH and detect when a cow experienced SARA. We used a positioning system to infer an animal’s activity based on its position in relation to specific elements in the barn (feeder, resting area, and alleys). We tested several ML algorithms: K Nearest Neighbours for Regression (KNNR); Decision Tree for Regression (DTR); MultiLayer Perceptron (MLP); Long Short-Term Memory (LSTM); and an algorithm where activity is assumed to be similar from one day to the next. First, we developed ML models to predict activity on a given day from the previous 24 h, considering all cows together. Then, we calculated the error between observed and predicted values for a given cow. Finally, we compared the error to a threshold chosen to optimise the distinction between normal and abnormal values. KNNR performed best, detecting 83% of SARA cases (true-positives), but it also produced 66% of false-positives, which limits its use in practice. In conclusion, ML can help detect anomalies in behaviour. Further improvements could probably be obtained by applying ML on very large datasets at animal rather than group level.

Behaviour classification of extensively grazed sheep using machine learning

The application of accelerometer sensors for automated animal behaviour monitoring is becoming increasingly common. Despite the rapid growth of research in this area, there is little consensus on the most appropriate method of data summation and analysis. The objective of this current study was to explore feature creation and machine learning (ML) algorithm options to provide the most accurate behavioural classification from an ear-borne accelerometer in extensively grazed sheep. Nineteen derived movement features, three epochs (5, 10 and 30 s) and four ML-algorithms (Classification and Regression Trees (CART), Linear Kernel Support-Vector Machines (SVM), Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA)) were assessed. Behaviour classification was also evaluated using three different ethograms, including detection of (i) grazing, lying, standing, walking; (ii) active and inactive behaviour; and (iii) body posture. Detection of the four mutually-exclusive behaviours (grazing, lying, standing and walking) was most accurately performed using a 10 s epoch by an SVM (76.9%). Activity was most accurately detected using a 30 s epoch by a CART (98.1%). LDA and a 30 s epoch was superior for detecting posture (90.6%). Differentiation relied on identification of disparities between behaviours rather than pattern recognition within a behaviour. The choice of epoch and ML algorithm will be dependent on the application purpose, with different combinations of each more accurate across the different ethograms. This study provides a crucial foundation for development of algorithms which can identify multiple behaviours in pasture-based sheep. This knowledge could be applied across a number of contexts, particularly at times of change in physiological or mental state e.g. during parturition or stress-inducing husbandry procedures.

227 Leveraging on high-throughput phenotyping technologies to optimize livestock genetic improvement and husbandry

Abstract The advent of fully automated data recording technologies and high-throughput phenotyping (HTP) systems has opened up a myriad of opportunities to advance breeding programs and livestock husbandry. Such technologies allow scoring large number of animals for novel phenotypes and indicator traits to boost genetic improvement, as well as for real-time monitoring of animal behavior and development for optimized management decisions. HTP tools include, for example, image analysis and computer vision, sensor technology for motion, sound and chemical composition, and spectroscopy. Applications span from health surveillance, precision nutrition, and control of meat and milk composition and quality. However, the application of HTP requires sophisticated statistical and computational approaches for efficient data management and appropriate data mining, as it involves large datasets with many covariates and complex relationships. In this talk we will discuss some of the challenges and potentials of HTP in livestock. Some examples to be presented include the utilization of automated feeders to record feed intake and to monitor feeding behavior in broilers, milk-spectra information to predict dairy cattle feed intake, and image analysis and computer vision to monitor growth and body condition in pigs and cattle. HTP and big data will become an essential component of modern livestock operations in the context of precision animal agriculture, boosting animal welfare, environmental footprint, and overall sustainability of animal production.

Animal species recognition in the wildlife based on muzzle and shape features using joint CNN

Monitoring of animal behavior in the wild supposes the reliable techniques for their species recognition using, mainly, visual data captured by camera traps. In this paper, we propose to extent Convolutional Neural Network (CNN) VGG by three branches, two of which are VGG16 for the muzzle and part of shape recognition and one is VGG19 for the whole shape recognition. A necessity of such branched CNN structure is caused by great variety of the animal poses fixed by a camera trap. Also, here we met with an objective problem of the unbalanced dataset due to different behavior of animals in nature. Preliminary categorization procedure of images helps to obtain better recognition results. Experiments were conducted using the dataset obtained from Ergaki national park, Krasnoyarsky Kray, Russia, 2012-2018. The joint CNN shows good accuracy results on the balanced dataset achieving 80.6% Top-1 and 94.1% Top-5, respectively. In the case of the unbalanced training dataset, we obtained 38.7% Top-1 and 54.8% Top-5 accuracy.

Design and test of a low-power acceleration sensor with Bluetooth Low Energy on ear tags for sow behaviour monitoring

Short battery life limits the applicability of wearable sensors in animal behaviour monitoring. This research developed an ear-attached acceleration sensor of low power consumption for sow welfare studies. The sensor weighed only 10.0 g and was composed of a 3-axis accelerometer, a Bluetooth Low Energy module, a printed circuit board antenna, and a 3-V, 230-mAh CR2032 battery. It entered into a sleeping state when no activities of the monitored sow were detected, to conserve battery energy. In the sleeping state, the sensor still read data from the accelerometer at 1 Hz, but did not transfer data wirelessly to a receiver if a first-order difference that differentiated sow resting from its other activities was below a defined threshold. Laboratory and 8-sow field tests were conducted to assess the sensor performance in three operational modes (continuous, data grouping, and power saving). Results showed that there were three major behaviours of the sows (resting, moving, and eating) and the behaviour patterns differed from sow to sow. The ear-attached sensors could in principle be used to monitor and predict sow farrowing while not causing sow abnormal behaviours. In the continuous mode, data broadcasting consumed 97.6% energy and the battery life was only about 31 days. The average battery life could be extended to about 288 days in power saving mode when the first-order difference threshold was set at 0.1 g. Compared with reported similar acceleration sensors, this new sensor provided more than double the battery life while reducing the sensor mass by more than 70%.

Infrared technology for estrous detection in Chinchilla lanigera

Thermography is one of the most modern methods for non-invasive monitoring of animal behavior, health status and physiological state. The sparcity of scientific reports related to this subject in fur animals prompted the use of thermography in Chinchilla lanigera. The aim of the study was to evaluate the use of a thermal imaging camera to diagnose or confirm estrus in Chinchilla lanigera. The subjects included 100 female chinchillas from three different breeds, selected during ovulation after parturition, which were evaluated twice: measurements were taken over three successive days starting from the 2nd day post partum and again through the same time period 6 weeks later when the females were ready to mate. A picture of the external genital area was taken with a FLIR C2 thermal imaging camera and the temperature was determined and recorded from the selected area using the FLIR TOOLS computer program. There were distinct differences between chinchillas with and without a confirmed estrus. Chinchillas in estrus had ROI temperatures that were greater (2.33 °C) than non-estrous females. During estrus, the temperature of the external genital area clearly increases, which could be due to genital hyperemia. This fact can be used for non-invasive estrous diagnosis, which is especially important because of the absence of typical estrous behavior in this species. Monitoring the greater surface temperature of the vulval area allows determination of the optimal time to place a female with a male and expect effective copulation.

Integrating direct observation and GPS tracking to monitor animal behavior for resource management.

Monitoring the behavior of pack animals in protected areas informs management about use patterns and the potential associated negative impacts. However, systematic assessments of behavior are uncommon due to methodological and logistical constraints. This study integrated behavior mapping with GPS tracking, and applied behavior change point analysis, as an approach to monitor the behaviors of pack animals during overnight periods. The integrated approach identified multiple grazing patterns (i.e., locally intense grazing, ambulatory grazing) not feasible through a single methodology alone. Monitoring behavior and corresponding environmental conditions aid managers in implementing strategies designed to mitigate impacts associated with pack animals in natural areas. Results also contrast the influence of temporal scale on behavior segmentation to inform decisions for further monitoring and management of domestic animal use and impacts in natural areas. This integrated approach reduced time and logistical constraints of each method individually to promote ongoing monitoring and highlight how multiple management tactics could reduce impacts to sensitive habitats.

Monitoring trough visits of growing-finishing pigs with UHF-RFID

Automatic monitoring of animal feeding behaviour in commercial farms is desirable as it is an important indicator for the well-being and health of animals. Low-frequency and high-frequency radio frequency identification (RFID) systems have been tested for the detection of feeding visits of growing-finishing pigs, but the suitability of ultra-high frequency (UHF) RFID for this application has not yet been shown. Therefore, the objective of this study was the validation of a UHF-RFID system, consisting of a reader, antennas and passive transponder ear tags, for the monitoring of visits of growing-finishing pigs at a short trough for liquid feeding. Consequently, (1) two antenna variants (free-form and patch antennas) were tested at different levels of antenna output power, (2) two methods to determine a bout criterion for the creation of trough visits from the RFID registrations were compared, and (3) a comparison of the RFID data with reference data from video observation was carried out. The analysis showed that the reading area exceeded the trough especially in the variants with high output power. Thus, the evaluation of the first and second rate of change of the total daily number and total daily duration of visits led to higher values for the bout criterion (50 s for both antenna variants) than the evaluation of the mean absolute deviation between video and RFID data (20 s for the free-form antenna and 30 s for the patch antennas). The trough visits observed were detected best with the patch antennas at 25 dBm output power and a bout criterion of 30 s with regard to average precision (61.1%), and correlation between the video data and RFID visits (R2 = 0.87 for total number and 0.80 for total duration of visits). The average sensitivity of this variant was 49.7%, specificity 99.0% and accuracy 97.9%. The highest average sensitivity (79.7%) and a good correlation between video data and RFID visits (R2 = 0.78 for total number, 0.56 for total duration of visits with a bout criterion of 50 s) were measured with the free-form antenna at 26 dBm. In conclusion, UHF-RFID can be suitable for the monitoring of trough visits of growing-finishing pigs, but the effect of ear tissue on the performance of the UHF-RFID ear tags should be reduced by further development. In addition, further research should be carried out to evaluate the potential of this technology completely for animal behaviour monitoring.

Embedded neural network for real-time animal behavior classification

Recent biological studies have focused on understanding animal interactions and welfare. To help biologists to obtain animals’ behavior information, resources like wireless sensor networks are needed. Moreover, large amounts of obtained data have to be processed off-line in order to classify different behaviors. There are recent research projects focused on designing monitoring systems capable of measuring some animals’ parameters in order to recognize and monitor their gaits or behaviors. However, network unreliability and high power consumption have limited their applicability.In this work, we present an animal behavior recognition, classification and monitoring system based on a wireless sensor network and a smart collar device, provided with inertial sensors and an embedded multi-layer perceptron-based feed-forward neural network, to classify the different gaits or behaviors based on the collected information. In similar works, classification mechanisms are implemented in a server (or base station). The main novelty of this work is the full implementation of a reconfigurable neural network embedded into the animal’s collar, which allows a real-time behavior classification and enables its local storage in SD memory. Moreover, this approach reduces the amount of data transmitted to the base station (and its periodicity), achieving a significantly improving battery life. The system has been simulated and tested in a real scenario for three different horse gaits, using different heuristics and sensors to improve the accuracy of behavior recognition, achieving a maximum of 81%.

Direction-of-arrival estimation of animal vocalizations for monitoring animal behavior and improving estimates of abundance

Direction-of-arrival estimation of animal vocalizations for monitoring animal behavior and improving estimates of abundance

Evaluation of a laying-hen tracking algorithm based on a hybrid support vector machine.

BackgroundBehavior is an important indicator reflecting the welfare of animals. Manual analysis of video is the most commonly used method to study animal behavior. However, this approach is tedious and depends on a subjective judgment of the analysts. There is an urgent need for automatic identification of individual animals and automatic tracking is a fundamental part of the solution to this problem.ResultsIn this study, an algorithm based on a Hybrid Support Vector Machine (HSVM) was developed for the automated tracking of individual laying hens in a layer group. More than 500 h of video was conducted with laying hens raised under a floor system by using an experimental platform. The experimental results demonstrated that the HSVM tracker outperformed the Frag (fragment-based tracking method), the TLD (Tracking-Learning-Detection), the PLS (object tracking via partial least squares analysis), the MeanShift Algorithm, and the Particle Filter Algorithm based on their overlap rate and the average overlap rate.ConclusionsThe experimental results indicate that the HSVM tracker achieved better robustness and state-of-the-art performance in its ability to track individual laying hens than the other algorithms tested. It has potential for use in monitoring animal behavior under practical rearing conditions.

Application of GPS and GIS to study foraging behavior of dairy cattle

Aplicación de GPS y GIS para estudiar el comportamiento en pastoreo de vacas lecheras. Se evaluó el uso de los Sistemas de Posicionamiento Global (GPS) para discriminar las actividades de pastoreo de vacas lecheras. El estudio se realizó en la granja robotizada de la Estación Biológica WK Kellogg de la Universidad Estatal de Michigan, entre 11 y 20 de agosto, 2010. Mediante observación de las actividades de forrajeo en diferentes sitios, usando sensores remotos se llevó un registro de la actividad de 4 vacas lactantes Holstein (650 kg PV; 23 kg.día-1) equipadas con collares GPS, que registran la posición de la cabeza con sensores de movimiento para los ejes X y Y. Los GPS mostraron 82-86% de probabilidad que la estimación de las ubicaciones de los animales tuviesen un error de 7 m. Estos datos sugirieron que las vacas permanecieron en las pasturas la mayor parte del tiempo (94,6±0,92%) y dentro del establo sólo un 5,4% (±0,92) del tiempo. Cuando las vacas estaban en las pasturas, la mayor parte del tiempo la dedicaron a pastoreo (51%); otra parte se dedicó a reposo (43%) y 6% a traslado. El ganado viajó en promedio 3385±712 m por día. En días de temperatura baja la actividad principal de las vacas fue el pastoreo (92%) y en días de media y alta temperatura el pastoreo fue sólo 62,6 y 59,4%, respectivamente. Por el contrario, el reposo fue la actividad más importante bajo media y alta temperatura (33,6 y 31,8%, respectivamente). Se demostró la utilidad de la teledetección y los GPS para monitorear el comportamiento animal.

Factors influencing interactions in zoos: animal-keeper relationship, animal-public interactions and solitary animals groups

Interactions that animals experience can have a significant influence on their health and welfare. These interactions can occur between animals themselves, but also between animals and keepers, and animals and the public. Human and non-human animals come into contact with each other in a variety of settings, and wherever there is contact there is the opportunity for interaction to take place. Interaction with companion animals are well known, but human–animal interaction (HAR) (Hosey, 2008) also occurs in the context of farms (Hemsworth and Gonyou, 1997; Hemsworth, 2003), laboratories (Chang and Hart, 2002), zoos (Kreger and Mench, 1995) and even the wild (e.g. Cassini, 2001). This project proposes a permanent monitoring scheme to record animal-human interactions and animal-animal interactions in zoos. This will be accompanied by a survey of animal personality for welfare, husbandry, breeding programs and reintroduction purposes. The pilot project is currently based on direct monitoring of animal behaviour, use of time lapse cameras and animal personality questionnaires completed by experienced keepers. The goal of this project is to create a network between zoos to explore the aforementioned interactions to produce husbandry protocols and explore personality and behavioural traits in multiple species. We present provisional data regarding polar bear (Fasano Zoosafari, Italy), Sumatran tigers, Amur tigers and Asiatic lion (ZSL London and Whipsnade zoo) interactions with humans and conspecifics. This data is collected across a broad range of environmental conditions and outlines the monitoring protocols developed to collect this data. The first year data show the great adaptability of these species to ex situ environments, low or absent negative impact of visitors’ presence and the relevance of individual personality in these interactions.

Individual differences in activity and habitat selection of juvenile queen conch evaluated using acceleration biologgers

Fine-scale differences in behaviour and habitat use have important ecological impli- cations, but have rarely been examined in marine gastropods. We used tri-axial accelerometer loggers to estimate activity levels and movement patterns of the juvenile queen conch Lobatus gigas (n = 11) in 2 habitat types in Eleuthera, The Bahamas. In 2 manipulations in nearshore areas, queen conchs were equipped with accelerometers and released in adjacent coral rubble or sea- grass habitats. Queen conchs were located approximately every 6 h during daylight by snorkel- ing, to measure individual differences in linear distance moved, and after 24 h they were relocated to an alternate habitat (24 h in each habitat). We found significant inter-individual variability in activity levels, but more consistent levels of activity between the 2 habitat types within individual queen conchs. Four (36%) of the individuals placed in seagrass moved back to the adjacent coral rubble habitat, suggesting selectivity for coral rubble. Individuals showed variable behavioural responses when relocated to the less preferable seagrass habitat, which may be related to differ- ing stress-coping styles. Our results suggest that behavioural variability between individuals may be an important factor driving movement and habitat use in queen conch and, potentially, their susceptibility to human stressors. This study provides evidence of diverse behavioural (activity) patterns and habitat selectivity in a marine gastropod and highlights the utility of accelero meter biologgers for continuously monitoring animal behaviour in the wild.