Give the machine a hand: A Boolean time‐based decision‐tree template for rapidly finding animal behaviours in multisensor data

Landing force reveals new form of motion-induced sound camouflage in a wild predator

Predator-prey arms races have led to the evolution of finely tuned disguise strategies. While the theoretical benefits of predator camouflage are well established, no study has yet been able to quantify its consequences for hunting success in natural conditions. We used high-resolution movement data to quantify how barn owls (Tyto alba) conceal their approach when using a sit-and-wait strategy. We hypothesized that hunting barn owls would modulate their landing force, potentially reducing noise levels in the vicinity of prey. Analysing 87,957 landings by 163 individuals equipped with GPS tags and accelerometers, we show that barn owls reduce their landing force as they approach their prey, and that landing force predicts the success of the following hunting attempt. Landing force also varied with the substrate, being lowest on man-made poles in field boundaries. The physical environment, therefore, affects the capacity for sound camouflage, providing an unexpected link between predator-prey interactions and land use. Finally, hunting strike forces in barn owls were the highest recorded in any bird, relative to body mass, highlighting the range of selective pressures that act on landings and the capacity of these predators to modulate their landing force. Overall, our results provide the first measurements of landing force in a wild setting, revealing a new form of motion-induced sound camouflage and its link to hunting success.

Inertial Navigation on Extremely Resource-Constrained Platforms: Methods, Opportunities and Challenges

Inertial navigation provides a small footprint, low-power, and low-cost pathway for localization in GPS-denied environments on extremely resource-constrained Internet-of-Things (IoT) platforms. Traditionally, application-specific heuristics and physics-based kinematic models are used to mitigate the curse of drift in inertial odometry. These techniques, albeit lightweight, fail to handle domain shifts and environmental non-linearities. Recently, deep neural-inertial sequence learning has shown superior odometric resolution in capturing non-linear motion dynamics without human knowledge over heuristic-based methods. These AI-based techniques are data-hungry, suffer from excessive resource usage, and cannot guarantee following the underlying system physics. This paper highlights the unique methods, opportunities, and challenges in porting real-time AI-enhanced inertial navigation algorithms onto IoT platforms. First, we discuss how platform-aware neural architecture search coupled with ultra-lightweight model backbones can yield neural-inertial odometry models that are 31-134× smaller yet achieve or exceed the localization resolution of state-of-the-art AI-enhanced techniques. The framework can generate models suitable for locating humans, animals, underwater sensors, aerial vehicles, and precision robots. Next, we showcase how techniques from neurosymbolic AI can yield physics-informed and interpretable neural-inertial navigation models. Afterward, we present opportunities for fine-tuning pre-trained odometry models in a new domain with as little as 1 minute of labeled data, while discussing inexpensive data collection and labeling techniques. Finally, we identify several open research challenges that demand careful consideration moving forward.

Highlighting when animals expend excessive energy for travel using Dynamic Body Acceleration

Travel represents a major cost for many animals so there should be selection pressure for it to be efficient – at minimum cost. However, animals sometimes exceed minimum travel costs for reasons that must be correspondingly important. We use Dynamic Body Acceleration (DBA), an acceleration-based metric, as a proxy for movement-based power, in tandem with vertical velocity (rate of change in depth) in a shark (Rhincodon typus) to derive the minimum estimated power required to swim at defined vertical velocities. We show how subtraction of measured DBA from the estimated minimum power for any given vertical velocity provides a “proxy for power above minimum” metric (PPA_min), highlighting when these animals travel above minimum power. We suggest that the adoption of this metric across species has value in identifying where and when animals are subject to compelling conditions that lead them to deviate from ostensibly judicious energy expenditure.

•

Plots of vertical speed vs DBA in sharks show swimming with minimum power

•

DBA values above this minimum indicate higher speeds or increases in drag

•

Linked to space use, this can identify regions and times of excess power use

TinyOdom: Hardware-Aware Efficient Neural Inertial Navigation

Deep inertial sequence learning has shown promising odometric resolution over model-based approaches for trajectory estimation in GPS-denied environments. However, existing neural inertial dead-reckoning frameworks are not suitable for real-time deployment on ultra-resource-constrained (URC) devices due to substantial memory, power, and compute bounds. Current deep inertial odometry techniques also suffer from gravity pollution, high-frequency inertial disturbances, varying sensor orientation, heading rate singularity, and failure in altitude estimation. In this paper, we introduce TinyOdom, a framework for training and deploying neural inertial models on URC hardware. TinyOdom exploits hardware and quantization-aware Bayesian neural architecture search (NAS) and a temporal convolutional network (TCN) backbone to train lightweight models targetted towards URC devices. In addition, we propose a magnetometer, physics, and velocity-centric sequence learning formulation robust to preceding inertial perturbations. We also expand 2D sequence learning to 3D using a model-free barometric g-h filter robust to inertial and environmental variations. We evaluate TinyOdom for a wide spectrum of inertial odometry applications and target hardware against competing methods. Specifically, we consider four applications: pedestrian, animal, aerial, and underwater vehicle dead-reckoning. Across different applications, TinyOdom reduces the size of neural inertial models by 31× to 134× with 2.5m to 12m error in 60 seconds, enabling the direct deployment of models on URC devices while still maintaining or exceeding the localization resolution over the state-of-the-art. The proposed barometric filter tracks altitude within ±0.1m and is robust to inertial disturbances and ambient dynamics. Finally, our ablation study shows that the introduced magnetometer, physics, and velocity-centric sequence learning formulation significantly improve localization performance even with notably lightweight models.

Decision rules for determining terrestrial movement and the consequences for filtering high-resolution global positioning system tracks: a case study using the African lion (Panthera leo)

The combined use of global positioning system (GPS) technology and motion sensors within the discipline of movement ecology has increased over recent years. This is particularly the case for instrumented wildlife, with many studies now opting to record parameters at high (infra-second) sampling frequencies. However, the detail with which GPS loggers can elucidate fine-scale movement depends on the precision and accuracy of fixes, with accuracy being affected by signal reception. We hypothesized that animal behaviour was the main factor affecting fix inaccuracy, with inherent GPS positional noise (jitter) being most apparent during GPS fixes for non-moving locations, thereby producing disproportionate error during rest periods. A movement-verified filtering (MVF) protocol was constructed to compare GPS-derived speed data with dynamic body acceleration, to provide a computationally quick method for identifying genuine travelling movement. This method was tested on 11 free-ranging lions (Panthera leo) fitted with collar-mounted GPS units and tri-axial motion sensors recording at 1 and 40 Hz, respectively. The findings support the hypothesis and show that distance moved estimates were, on average, overestimated by greater than 80% prior to GPS screening. We present the conceptual and mathematical protocols for screening fix inaccuracy within high-resolution GPS datasets and demonstrate the importance that MVF has for avoiding inaccurate and biased estimates of movement.

How often should dead-reckoned animal movement paths be corrected for drift?

BACKGROUND

Understanding what animals do in time and space is important for a range of ecological questions, however accurate estimates of how animals use space is challenging. Within the use of animal-attached tags, radio telemetry (including the Global Positioning System, 'GPS') is typically used to verify an animal's location periodically. Straight lines are typically drawn between these 'Verified Positions' ('VPs') so the interpolation of space-use is limited by the temporal and spatial resolution of the system's measurement. As such, parameters such as route-taken and distance travelled can be poorly represented when using VP systems alone. Dead-reckoning has been suggested as a technique to improve the accuracy and resolution of reconstructed movement paths, whilst maximising battery life of VP systems. This typically involves deriving travel vectors from motion sensor systems and periodically correcting path dimensions for drift with simultaneously deployed VP systems. How often paths should be corrected for drift, however, has remained unclear.

METHODS AND RESULTS

Here, we review the utility of dead-reckoning across four contrasting model species using different forms of locomotion (the African lion Panthera leo, the red-tailed tropicbird Phaethon rubricauda, the Magellanic penguin Spheniscus magellanicus, and the imperial cormorant Leucocarbo atriceps). Simulations were performed to examine the extent of dead-reckoning error, relative to VPs, as a function of Verified Position correction (VP correction) rate and the effect of this on estimates of distance moved. Dead-reckoning error was greatest for animals travelling within air and water. We demonstrate how sources of measurement error can arise within VP-corrected dead-reckoned tracks and propose advancements to this procedure to maximise dead-reckoning accuracy.

CONCLUSIONS

We review the utility of VP-corrected dead-reckoning according to movement type and consider a range of ecological questions that would benefit from dead-reckoning, primarily concerning animal-barrier interactions and foraging strategies.

Animal tag technology keeps coming of age: an engineering perspective

Animal-borne tags (biologgers) have now become extremely sophisticated, recording data from multiple sensors at high frequencies for long periods and, as such, have become a powerful tool for behavioural ecologists and physiologists studying wild animals. But the design and implementation of these tags is not trivial because engineers have to maximize performance and ability to function under onerous conditions while minimizing tag mass and volume (footprint) to maximize the wellbeing of the animal carriers. We present some of the major issues faced by tag engineers and show how tag designers must accept compromises while maintaining systems that can answer the questions being posed. We also argue that basic understanding of engineering issues in tag design by biologists will help feedback to engineers to better tag construction but also reduce the likelihood that tag-deploying biologists will misunderstand their own results. Finally, we suggest that proper consideration of conventional technology together with new approaches will lead to further step changes in our understanding of wild-animal biology using smart tags. This article is part of the theme issue 'Measuring physiology in free-living animals (Part II)'.

Limitations of using surrogates for behaviour classification of accelerometer data: refining methods using random forest models in Caprids

BACKGROUND

Animal-attached devices can be used on cryptic species to measure their movement and behaviour, enabling unprecedented insights into fundamental aspects of animal ecology and behaviour. However, direct observations of subjects are often still necessary to translate biologging data accurately into meaningful behaviours. As many elusive species cannot easily be observed in the wild, captive or domestic surrogates are typically used to calibrate data from devices. However, the utility of this approach remains equivocal.

METHODS

Here, we assess the validity of using captive conspecifics, and phylogenetically-similar domesticated counterparts (surrogate species) for calibrating behaviour classification. Tri-axial accelerometers and tri-axial magnetometers were used with behavioural observations to build random forest models to predict the behaviours. We applied these methods using captive Alpine ibex (Capra ibex) and a domestic counterpart, pygmy goats (Capra aegagrus hircus), to predict the behaviour including terrain slope for locomotion behaviours of captive Alpine ibex.

RESULTS

Behavioural classification of captive Alpine ibex and domestic pygmy goats was highly accurate (> 98%). Model performance was reduced when using data split per individual, i.e., classifying behaviour of individuals not used to train models (mean ± sd = 56.1 ± 11%). Behavioural classifications using domestic counterparts, i.e., pygmy goat observations to predict ibex behaviour, however, were not sufficient to predict all behaviours of a phylogenetically similar species accurately (> 55%).

CONCLUSIONS

We demonstrate methods to refine the use of random forest models to classify behaviours of both captive and free-living animal species. We suggest there are two main reasons for reduced accuracy when using a domestic counterpart to predict the behaviour of a wild species in captivity; domestication leading to morphological differences and the terrain of the environment in which the animals were observed. We also identify limitations when behaviour is predicted in individuals that are not used to train models. Our results demonstrate that biologging device calibration needs to be conducted using: (i) with similar conspecifics, and (ii) in an area where they can perform behaviours on terrain that reflects that of species in the wild.

Quantifying the frequency and volume of urine deposition by grazing sheep using tri-axial accelerometers

Urine patches deposited in pasture by grazing animals are sites of reactive nitrogen (N) loss to the environment due to high concentrations of N exceeding pasture uptake requirements. In order to upscale N losses from the urine patch, several urination parameters are required, including where, when and how often urination events occur as well as the volume and chemical composition. There are limited data available in this respect, especially for sheep. Here, we seek to address this knowledge gap by using non-invasive sensor-based technology (accelerometers) on ewes grazing in situ, using a Boolean algorithm to detect urination events in the accelerometer signal. We conducted an initial study with penned Welsh Mountain ewes (n = 5), with accelerometers attached to the hind, to derive urine flow rate and to determine whether urine volume could be estimated from ewe squat time. Then accelerometers attached to the hind of Welsh Mountain ewes (n = 30 at each site) were used to investigate the frequency of sheep urination events (n = 35 946) whilst grazing two extensively managed upland pastures (semi-improved and unimproved) across two seasons (spring and autumn) at each site (35-40 days each). Sheep urinated at a frequency of 10.2 ± 0.2 and 8.1 ± 0.3 times per day in the spring and autumn, respectively, while grazing the semi-improved pasture. Urination frequency was greater (19.0 ± 0.4 and 15.3 ± 0.3 times per day in the spring and autumn, respectively) in the unimproved pasture. Ewe squat duration could be reliably used to predict the volume of urine deposited per event and was thus used to estimate mean daily urine production volumes. Sheep urinated at a rate of 16.6 mL/s and, across the entire dataset, sheep squatted for an average of 9.62 ± 0.03 s per squatting event, producing an estimated average individual urine event volume of 159 ± 1 mL (n = 35 946 events), ranging between 17 and 745 mL (for squat durations of 1 to 45 s). The estimated mean daily urine volume was 2.15 ± 0.04 L (n = 2 669 days) across the entire dataset. The data will be useful for modelling studies estimating N losses (e.g. ammonia (NH₃) volatilisation, nitrous oxide (N₂O) emission via nitrification and denitrification and nitrate (NO₃^-) leaching) from urine patches.

Less is more: On-board lossy compression of accelerometer data increases biologging capacity

GPS‐tracking devices have been used in combination with a wide range of additional sensors to study animal behaviour, physiology and interaction with their environment. Tri‐axial accelerometers allow researchers to remotely infer the behaviour of individuals, at all places and times. Collection of accelerometer data is relatively cheap in terms of energy usage, but the amount of raw data collected generally requires much storage space and is particularly demanding in terms of energy needed for data transmission.

Here, we propose compressing the raw accelerometer (ACC) data into summary statistics within the tracking device (before transmission) to reduce data size, as a means to overcome limitations in storage and energy capacity.

We explored this type of lossy data compression in the accelerometer data of tagged Bewick's swans Cygnus columbianus bewickii collected in spring 2017. Using software settings in which bouts of 2 s of both raw ACC data and summary statistics were collected in parallel but with different bout intervals to keep total data size comparable, we created the opportunity for a direct comparison of time budgets derived by the two data collection methods.

We found that the data compression in our case yielded a six times reduction in data size per bout, and concurrent, similar decreases in storage and energy use of the device. We show that with the same accuracy of the behavioural classification, the freed memory and energy of the device can be used to increase the monitoring effort, resulting in a more detailed representation of the individuals’ time budget. Rare and/or short behaviours, such as daily roost flights, were picked up significantly more when collecting summary statistics instead of raw ACC data (but note differences in sampling rate). Such level of detail can be of essential importance, for instance to make a reliable estimate of the energy budgets of individuals.

In conclusion, we argue that this type of lossy data compression can be a well‐considered choice in study situations where limitations in energy and storage space of the device pose a problem. Ultimately, these developments can allow for long‐term and nearly continuous remote monitoring of the behaviour of free‐ranging animals.

Physical limits of flight performance in the heaviest soaring bird

Flapping flight is extremely costly for large birds, yet little is known about the conditions that force them to flap. We attached custom-made “flight recorders” to Andean condors, the world’s heaviest soaring birds, documenting every single wingbeat and when and how individuals gained altitude. Remarkably, condors flapped for only 1% of their flight time, specifically during takeoff and when close to the ground. This is particularly striking as the birds were immature. Thus, our results demonstrate that even inexperienced birds can cover vast distances over land without flapping. Overall, this can help explain how extinct birds with twice the wingspan of condors could have flown.

Flight costs are predicted to vary with environmental conditions, and this should ultimately determine the movement capacity and distributions of large soaring birds. Despite this, little is known about how flight effort varies with environmental parameters. We deployed bio-logging devices on the world’s heaviest soaring bird, the Andean condor (Vultur gryphus), to assess the extent to which these birds can operate without resorting to powered flight. Our records of individual wingbeats in >216 h of flight show that condors can sustain soaring across a wide range of wind and thermal conditions, flapping for only 1% of their flight time. This is among the very lowest estimated movement costs in vertebrates. One bird even flew for >5 h without flapping, covering ∼172 km. Overall, > 75% of flapping flight was associated with takeoffs. Movement between weak thermal updrafts at the start of the day also imposed a metabolic cost, with birds flapping toward the end of glides to reach ephemeral thermal updrafts. Nonetheless, the investment required was still remarkably low, and even in winter conditions with weak thermals, condors are only predicted to flap for ∼2 s per kilometer. Therefore, the overall flight effort in the largest soaring birds appears to be constrained by the requirements for takeoff.

A new direction for differentiating animal activity based on measuring angular velocity about the yaw axis

The use of animal-attached data loggers to quantify animal movement has increased in popularity and application in recent years. High-resolution tri-axial acceleration and magnetometry measurements have been fundamental in elucidating fine-scale animal movements, providing information on posture, traveling speed, energy expenditure, and associated behavioral patterns. Heading is a key variable obtained from the tandem use of magnetometers and accelerometers, although few field investigations have explored fine-scale changes in heading to elucidate differences in animal activity (beyond the notable exceptions of dead-reckoning).This paper provides an overview of the value and use of animal heading and a prime derivative, angular velocity about the yaw axis, as an important element for assessing activity extent with potential to allude to behaviors, using "free-ranging" Loggerhead turtles (Caretta caretta) as a model species.We also demonstrate the value of yaw rotation for assessing activity extent, which varies over the time scales considered and show that various scales of body rotation, particularly rate of change of yaw, can help resolve differences between fine-scale behavior-specific movements. For example, oscillating yaw movements about a central point of the body's arc implies bouts of foraging, while unusual circling behavior, indicative of conspecific interactions, could be identified from complete revolutions of the longitudinal axis.We believe this approach should help identification of behaviors and "space-state" approaches to enhance our interpretation of behavior-based movements, particularly in scenarios where acceleration metrics have limited value, such as for slow-moving animals.

An "orientation sphere" visualization for examining animal head movements

Animal behavior is elicited, in part, in response to external conditions, but understanding how animals perceive the environment and make the decisions that bring about these behavioral responses is challenging.Animal heads often move during specific behaviors and, additionally, typically have sensory systems (notably vision, smell, and hearing) sampling in defined arcs (normally to the front of their heads). As such, head-mounted electronic sensors consisting of accelerometers and magnetometers, which can be used to determine the movement and directionality of animal heads (where head "movement" is defined here as changes in heading [azimuth] and/or pitch [elevation angle]), can potentially provide information both on behaviors in general and also clarify which parts of the environment the animals might be prioritizing ("environmental framing").We propose a new approach to visualize the data of such head-mounted tags that combines the instantaneous outputs of head heading and pitch in a single intuitive spherical plot. This sphere has magnetic heading denoted by "longitude" position and head pitch by "latitude" on this "orientation sphere" (O-sphere).We construct the O-sphere for the head rotations of a number of vertebrates with contrasting body shape and ecology (oryx, sheep, tortoises, and turtles), illustrating various behaviors, including foraging, walking, and environmental scanning. We also propose correcting head orientations for body orientations to highlight specific heading-independent head rotation, and propose the derivation of O-sphere-metrics, such as angular speed across the sphere. This should help identify the functions of various head behaviors.Visualizations of the O-sphere provide an intuitive representation of animal behavior manifest via head orientation and rotation. This has ramifications for quantifying and understanding behaviors ranging from navigation through vigilance to feeding and, when used in tandem with body movement, should provide an important link between perception of the environment and response to it in free-ranging animals.

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.

Monitoring canid scent marking in space and time using a biologging and machine learning approach

For canid species, scent marking plays a critical role in territoriality, social dynamics, and reproduction. However, due in part to human dependence on vision as our primary sensory modality, research on olfactory communication is hampered by a lack of tractable methods. In this study, we leverage a powerful biologging approach, using accelerometers in concert with GPS loggers to monitor and describe scent-marking events in time and space. We performed a validation experiment with domestic dogs, monitoring them by video concurrently with the novel biologging approach. We attached an accelerometer to the pelvis of 31 dogs (19 males and 12 females), detecting raised-leg and squat posture urinations by monitoring the change in device orientation. We then deployed this technique to describe the scent marking activity of 3 guardian dogs as they defend livestock from coyote depredation in California, providing an example use-case for the technique. During validation, the algorithm correctly classified 92% of accelerometer readings. High performance was partly due to the conspicuous signatures of archetypal raised-leg postures in the accelerometer data. Accuracy did not vary with the weight, age, and sex of the dogs, resulting in a method that is broadly applicable across canid species’ morphologies. We also used models trained on each individual to detect scent marking of others to emulate the use of captive surrogates for model training. We observed no relationship between the similarity in body weight between the dog pairs and the overall accuracy of predictions, although models performed best when trained and tested on the same individual. We discuss how existing methods in the field of movement ecology can be extended to use this exciting new data type. This paper represents an important first step in opening new avenues of research by leveraging the power of modern-technologies and machine-learning to this field.

Optimizing the use of biologgers for movement ecology research

The paradigm-changing opportunities of biologging sensors for ecological research, especially movement ecology, are vast, but the crucial questions of how best to match the most appropriate sensors and sensor combinations to specific biological questions and how to analyse complex biologging data, are mostly ignored. Here, we fill this gap by reviewing how to optimize the use of biologging techniques to answer questions in movement ecology and synthesize this into an Integrated Biologging Framework (IBF). We highlight that multisensor approaches are a new frontier in biologging, while identifying current limitations and avenues for future development in sensor technology. We focus on the importance of efficient data exploration, and more advanced multidimensional visualization methods, combined with appropriate archiving and sharing approaches, to tackle the big data issues presented by biologging. We also discuss the challenges and opportunities in matching the peculiarities of specific sensor data to the statistical models used, highlighting at the same time the large advances which will be required in the latter to properly analyse biologging data. Taking advantage of the biologging revolution will require a large improvement in the theoretical and mathematical foundations of movement ecology, to include the rich set of high-frequency multivariate data, which greatly expand the fundamentally limited and coarse data that could be collected using location-only technology such as GPS. Equally important will be the establishment of multidisciplinary collaborations to catalyse the opportunities offered by current and future biologging technology. If this is achieved, clear potential exists for developing a vastly improved mechanistic understanding of animal movements and their roles in ecological processes and for building realistic predictive models.

Luck in Food Finding Affects Individual Performance and Population Trajectories

Energy harvesting by animals is important because it provides the power needed for all metabolic processes. Beyond this, efficient food finding enhances individual fitness [1] and population viability [2], although rates of energy accumulation are affected by the environment and food distribution. Typically, differences between individuals in the rate of food acquisition are attributed to varying competencies [3], even though food-encounter rates are known to be probabilistic [4]. We used animal-attached technology to quantify food intake in four disparate free-living vertebrates (condors, cheetahs, penguins, and sheep) and found that inter-individual variability depended critically on the probability of food encounter. We modeled this to reveal that animals taking rarer food, such as apex predators and scavengers, are particularly susceptible to breeding failure because this variability results in larger proportions of the population failing to accrue the necessary resources for their young before they starve and because even small changes in food abundance can affect this variability disproportionately. A test of our model on wild animals indicated why Magellanic penguins have a stable population while the congeneric African penguin population has declined for decades. We suggest that such models predicting probabilistic ruin can help predict the fortunes of species operating under globally changing conditions.