Animal-borne Sensor Research Articles

SensorDrop: A system to remotely detach individual sensors from wildlife tracking collars.

The growing diversity of animal-borne sensor types is revolutionizing our understanding of wildlife biology. For example, researcher-developed sensors, such as audio and video loggers, are being increasingly attached to wildlife tracking collars to provide insights into a range of topics from species interactions to physiology. However, such devices are often prohibitively power-intensive, relative to conventional wildlife collar sensors, and their retrieval without compromising long-term data collection and animal welfare remains a challenge. We present an open-source system (SensorDrop) for remotely detaching individual sensors from wildlife collars. SensorDrop facilitates the retrieval of power-intensive sensors while leaving non-resource-intensive sensors intact on animals. SensorDrop systems can be made using commercially available components and are a fraction of the cost of other timed drop-off devices that detach full wildlife tracking collars. From 2021 to 2022, eight SensorDrop units were successfully deployed on free-ranging African wild dog packs in the Okavango Delta as part of audio-accelerometer sensor bundles attached to wildlife collars. All SensorDrop units detached after 2-3 weeks and facilitated the collection of audio and accelerometer data while leaving wildlife GPS collars intact to continue collecting locational data (>1 year), critical for long-term conservation population monitoring in the region. SensorDrop offers a low-cost method to remotely detach and retrieve individual sensors from wildlife collars. By selectively detaching battery-depleted sensors, SensorDrop maximizes the amount of data collected per wildlife collar deployment and mitigates ethical concerns on animal rehandling. SensorDrop adds to the growing body of open-source animal-borne technologies being utilized by wildlife researchers to innovate and expand upon data collection practices and supports the continued ethical use of novel technologies within wildlife studies.

Supervised versus unsupervised approaches to classification of accelerometry data.

Sophisticated animal-borne sensor systems are increasingly providing novel insight into how animals behave and move. Despite their widespread use in ecology, the diversity and expanding quality and quantity of data they produce have created a need for robust analytical methods for biological interpretation. Machine learning tools are often used to meet this need. However, their relative effectiveness is not well known and, in the case of unsupervised tools, given that they do not use validation data, their accuracy can be difficult to assess. We evaluated the effectiveness of supervised (n = 6), semi-supervised (n = 1), and unsupervised (n = 2) approaches to analyzing accelerometry data collected from critically endangered California condors (Gymnogyps californianus). Unsupervised K-means and EM (expectation-maximization) clustering approaches performed poorly, with adequate classification accuracies of <0.8 but very low values for kappa statistics (range: -0.02 to 0.06). The semi-supervised nearest mean classifier was moderately effective at classification, with an overall classification accuracy of 0.61 but effective classification only of two of the four behavioral classes. Supervised random forest (RF) and k-nearest neighbor (kNN) machine learning models were most effective at classification across all behavior types, with overall accuracies >0.81. Kappa statistics were also highest for RF and kNN, in most cases substantially greater than for other modeling approaches. Unsupervised modeling, which is commonly used for the classification of a priori-defined behaviors in telemetry data, can provide useful information but likely is instead better suited to post hoc definition of generalized behavioral states. This work also shows the potential for substantial variation in classification accuracy among different machine learning approaches and among different metrics of accuracy. As such, when analyzing biotelemetry data, best practices appear to call for the evaluation of several machine learning techniques and several measures of accuracy for each dataset under consideration.

Mobilizing Animal Movement Data: API use and the Movebank platform

Movebank, a global platform for animal tracking and other animal-borne sensor data, is used by over 3,000 researchers globally to harmonize, archive and share nearly 3 billion animal occurrence records and more than 3 billion other animal-borne sensor measurements that document the movements and behavior of over 1,000 species. Movebank’s publicly described data model (Kranstauber et al. 2011), vocabulary and application programming interfaces (APIs) provide services for users to automate data import and retrieval. Near-live data feeds are maintained in cooperation with over 20 manufacturers of animal-borne sensors, who provide data in agreed-upon formats for accurate data import. Data acquisition by API complies with public or controlled-access sharing settings, defined within the database by data owners. The Environmental Data Automated Track Annotation System (EnvDATA, Dodge et al. 2013) allows users to link animal tracking data with hundreds of environmental parameters from remote sensing and weather reanalysis products through the Movebank website, and offers an API for advanced users to automate the submission of annotation requests. Movebank's mobile apps, the Animal Tracker and Animal Tagger, use APIs to support reporting and monitoring while in the field, as well as communication with citizen scientists. The recently-launched MoveApps platform connects with Movebank data using an API to allow users to build, execute and share repeatable workflows for data exploration and analysis through a user-friendly interface. A new API, currently under development, will allow calls to retrieve data from Movebank reduced according to criteria defined by "reduction profiles", which can greatly reduce the volume of data transferred for many use cases. In addition to making this core set of Movebank services possible, Movebank's APIs enable the development of external applications, including the widely used R programming packages 'move' (Kranstauber et al. 2012) and 'ctmm' (Calabrese et al. 2016), and user-specific workflows to efficiently execute collaborative analyses and automate tasks such as syncing with local organizational and governmental websites and archives. The APIs also support large-scale data acquisition, including for projects under development to visualize, map and analyze bird migrations led by the British Trust for Ornithology, the coordinating organisation for European bird ringing (banding) schemes (EURING), Georgetown University, National Audubon Society, Smithsonian Institution and United Nations Convention on Migratory Species. Our API development is constrained by a lack of standardization in data reporting across animal-borne sensors and a need to ensure adequate communication with data users (e.g., how to properly interpret data; expectations for use and attribution) and data owners (e.g., who is using publicly-available data and how) when allowing automated data access. As interest in data linking, harvesting, mirroring and integration grows, we recognize needs to coordinate API development across animal tracking and biodiversity databases, and to develop a shared system for unique organism identifiers. Such a system would allow linking of information about individual animals within and across repositories and publications in order to recognize data for the same individuals across platforms, retain provenance and attribution information, and ensure beneficial and biologically meaningful data use.

Development of light-weight video-tracking technology for use in wildlife research: A case study on kangaroos

There have been significant advances in the development of animal-borne sensor technologies, or biologgers, in recent years. This has resulted in tremendous capacity for wildlife researchers to remotely collect physiological, behavioural and social data from wildlife in circumstances that were unthinkable just decades ago. While this technology can provide us with a unique insight into the “secret lives” of wild animals, there is a need to evaluate the utility of these new sensors versus traditional wildlife research methodologies, and to critically evaluate the integrity of the data collected by ensuring that these devices themselves do not alter the physiology or behaviour of the recipient animal. This paper reports on the development of a light weight “animal borne video and environmental data collection system” (AVED), which can be deployed on animals as small as 11 kg, whilst still meeting the desired 3% body weight threshold. This AVED (referred to as the “Kangaroo-cam”) simultaneously collects video footage and GPS location data for an average of 19 h. Kangaroo-cams were deployed on seven kangaroos as a proof of concept of their potential utility for the study of location specific behaviour and diet in a medium-sized terrestrial herbivore. Following device recovery and data processing, we were able to successfully score 83 foraging events which allowed us to determine diet based on visual identification (to the family level) of plants consumed. This approach could be further broadened to include a comparison of plant species consumed versus plant species encountered to provide a novel approach to diet selection analysis. When combined with GPS mapping of foraging locations, this approach would allow researchers to address questions on diet selection at both fine (within patch) and broad (habitat) spatial scales, overcoming some of the limitations of traditional diet selection methodologies. However, animal capture and collar deployment caused a significant elevation in stress hormone concentrations within the first 24 h post-capture, which highlighted the need to incorporate a time-delay capacity into these devices. We conclude the paper by reviewing recent advances in the development of AVED technology and providing suggestions for the improvement of this Kangaroo-cam device.

Bio-logging Data in Darwin Core: Use Cases

Animal-borne sensor data, along with other types of sensor-based observations, provide a growing volume and proportion of documentation about biodiversity. These data differ from the traditional specimen, sampling and human observation records for which the Taxonomic Database Working Group (TDWG) originally designed the Darwin Core standard. The original intention of the new TDWG Machine Observations Interest Group is to facilitate a body of work combining the informatics expertise of TDWG with that of subject matter experts to document best practice guidelines for applying Darwin Core to bio-logging datasets. This session offers the opportunity to walk through some of the use cases developed so far, including a terrestrial GPS tracking and acceleration dataset from Movebank and a marine acoustic telemetry dataset from the Ocean Tracking Network using stationary as well as mobile acoustic receivers. Through these examples, we will describe the strategy and rationale for the approaches taken to the application of Darwin Core using typical animal tracking scenarios laced with some of the common complexities in bio-logging and other types of machine-based biodiversity observations.

Reduced Energy and Memory Requirements by On-Board Behavior Classification for Animal-Borne Sensor Applications

The ability to study animal behavior is important in many fields of science, including behavioral ecology, conservation, and precision farming. These studies typically employ biotelemetry tags attached to animals that collect raw sensor data from tri-axial accelerometers and global positioning system modules. The lifespan of such tags is constrained by their power and memory usage, which are often limiting factors when performing behavioral studies for extended periods of time. This paper considers the power requirement and memory usage benefits of performing statistical behavior classification on the tag itself, as opposed to at a receiver station after raw data transmission. Experiments using specially designed low-power biotelemetry sensors demonstrated a 27-fold reduction in energy consumption and a 469-fold reduction in memory usage when classification was performed on the tag, rather than after raw data transmission. By performing on-board statistical behavior classification, both the power requirements and the memory usage are drastically reduced, thereby prolonging the lifespan of the tag.

Design and adaptation of a folded split ring resonator antenna for use in an animal-borne sensor

We present the design, optimisation and practical evaluation of a folded split ring resonator (FSRR) antenna for the purpose of radio communication with an animal-borne sensor. We show that the measurements agree with the simulated results and that we are able to produce an electrically small antenna with low mismatch, high radiation efficiency and a quasi-isotropic radiation pattern. We then adapt the topology of the design from a circular to a rectangular shape, to completely fit inside the sensor enclosure. A quasi-isotropic pattern is maintained as well as low mismatch by appropriate tuning. There is a decrease in radiation efficiency which may be countered by a thinner substrate and retuning. We conclude that the adapted FSRR antenna is a suitable design for our application.

Automated proximity sensing in small vertebrates: design of miniaturized sensor nodes and first field tests in bats.

Social evolution has led to a stunning diversity of complex social behavior, in particular in vertebrate taxa. Thorough documentation of social interactions is crucial to study the causes and consequences of sociality in gregarious animals. Wireless digital transceivers represent a promising tool to revolutionize data collection for the study of social interactions in terms of the degree of automation, data quantity, and quality. Unfortunately, devices for automated proximity sensing via direct communication among animal‐borne sensors are usually heavy and do not allow for the investigation of small animal species, which represent the majority of avian and mammalian taxa. We present a lightweight animal‐borne sensor node that is built from commercially available components and uses a sophisticated scheme for energy‐efficient communication, with high sampling rates at relatively low power consumption. We demonstrate the basic functionality of the sensor node under laboratory conditions and its applicability for the study of social interactions among free‐ranging animals. The first field tests were performed on two species of bats in temperate and tropical ecosystems. At <2 g, this sensor node is light enough to observe a broad spectrum of taxa including small vertebrates. Given our specifications, the system was especially sensitive to changes in distance within the short range (up to a distance of 4 m between tags). High spatial resolution at short distances enables the evaluation of interactions among individuals at a fine scale and the investigation of close contacts. This technology opens new avenues of research, allowing detailed investigation of events associated with social contact, such as mating behavior, pathogen transmission, social learning, and resource sharing. Social behavior that is not easily observed becomes observable, for example, in animals living in burrows or in nocturnal animals. A switch from traditional methods to the application of digital transceiver chips in proximity sensing offers numerous advantages in addition to an enormous increase in data quality and quantity. For future applications, the platform allows for the integration of additional sensors that may collect physiological or environmental data. Such information complements social network studies and may allow for a deeper understanding of animal ecology and social behavior.

Accounting for Location Error in Kalman Filters: Integrating Animal Borne Sensor Data into Assimilation Schemes

Data assimilation is a crucial aspect of modern oceanography. It allows the future forecasting and backward smoothing of ocean state from the noisy observations. Statistical methods are employed to perform these tasks and are often based on or related to the Kalman filter. Typically Kalman filters assumes that the locations associated with observations are known with certainty. This is reasonable for typical oceanographic measurement methods. Recently, however an alternative and abundant source of data comes from the deployment of ocean sensors on marine animals. This source of data has some attractive properties: unlike traditional oceanographic collection platforms, it is relatively cheap to collect, plentiful, has multiple scientific uses and users, and samples areas of the ocean that are often difficult of costly to sample. However, inherent uncertainty in the location of the observations is a barrier to full utilisation of animal-borne sensor data in data-assimilation schemes. In this article we examine this issue and suggest a simple approximation to explicitly incorporate the location uncertainty, while staying in the scope of Kalman-filter-like methods. The approximation stems from a Taylor-series approximation to elements of the updating equation.