Study of animal movement: design and development of equipment
This project aims to answer questions about the fundamental biology of native animal species in a state of vulnerability. For this, we design and develop equipment, carry out field monitoring, analyze the data we extract from our campaigns and adapt movement models to the data obtained.About Us
We are a group composed of biologists, physicists and engineers, and we have been working in interdisciplinarity for 5 years with different animal species, different habitats and challenges. We conduct our research at the Bariloche Atomic Center, San Carlos de Bariloche, Río Negro, Argentina.…And how did we start?
Five years ago, I was developing theoretical models of animal movement under the supervision of Dr. Guillermo Abramson. Around the second half of my PhD, it was necessary for me and for research to start with real animal tracking. We started by monitoring the marsupial Dromiciops gliroids in its natural habitat. This work turned out to be more than arduous given that this animal lives in the forest where there is no GPS signal, that it weighs about 20 grams (which greatly limits the usable material) and that it is It is a nocturnal animal (with which monitoring must be carried out at night). We have developed our own antennas to set up receiving stations and placed transmitters on individuals of D. gliroids.
A team member located at one of the receiving stations in the forest monitors Dromiciops gliroides using radio telemetry techniques.
Our project continued to grow and many conservation groups asked us if we could develop monitoring systems that could be adapted to their context. We started to develop our own monitoring system including GPS, inertial sensors (accelerometer, gyroscope and magnetometer) and temperature sensor. On the one hand, we have grown in the research and development of our monitoring systems and, in addition, we have received the TinyML kits, which can also become monitoring systems. Now, with SenseCAP K1100, we could set up a connected network of devices that simultaneously monitor animals.Nowadays…
Thanks to these tracking systems (which we will detail in the next section), we can obtain the trajectories of the animals. In particular, we are now interested in monitoring a population of the tortoise Chelonoidis chilensis in its natural habitat, located in San Antonio Oeste, Río Negro, Argentina. This species is in a vulnerable state (IUCN 2014, https://www.iucnredlist.org/species/9007/12949680) and very little is known about its biology. We believe that by studying the interdisciplinary system (development and implementation of tracking devices, data analysis and development of theoretical models), we can learn more about this species. Knowing how animals move, monitoring their trajectories, finding their nests, studying how their behavior depends on external variables (such as temperature or wind, for example) are essential to establish guidelines that help in their conservation.
A team member holding an individual Chelonoidis chilensis in its natural habitat.
We use devices that contain GPS and inertial sensors to monitor animal movements and determine their activity based on different variables (time of day, weather conditions, etc.). These devices can vary in technology and performance, but must be light and small enough to meet the requirements of each biological system. With this, the possibilities are reduced. We will present below the 2 monitoring systems with which we work.
- Tracking system designed by our team. This device uses a CC1312R1F3RGZT microcontroller and contains GPS, inertial sensors, temperature sensors and a microphone (see circuit diagram “Animal monitoring system using the CC1312R1F3RGZT microcontroller” in “Schematics and Circuit Diagrams”). We include sample code for the IMU module transmitting and receiving data for this device (see “Sample code to transmit/receive IMU data” in the “Code” section). In this section, we also include some Python code “Graphical user interfaceto interactively display temperature and accelerometer measurements. Using the first prototype we developed for this device, we collected enough data to study the motion of the turtle C.chilensis. With these results, we submitted the manuscript “Study of the movement of the vulnerable turtle Chelonoidis chilensis followed by complementary techniques”.
Three-dimensional view of the monitoring device developed by our team. The dimensions of this monitoring unit are 63×48 mm.
During the next monitoring campaign (January 2023), we will use this device to track individuals and obtain new movement data. At the moment the data is stored in a memory, so we don’t get the data until we retrieve the memory. Our goal is to use long-range, low-power communication protocols like LoRa to be able to see animal behavior in real time.
- TinyML + Edge Impulse kits. Our team is part of the TinyML network. Thanks to this spectacular initiative, we received 10 TinyML kits which consist, among other components, of the Nano 33 BLE Sense which contains a 9-axis inertial measurement unit (IMU) and temperature. Integrating GPS, battery and memory, we can use them as new animal monitoring devices. By using a sufficiently light and small monitoring device as well as data analysis and machine learning algorithms, it is possible to determine the behavior of individuals on the spot. Using TinyML kits and the Edge Impulse platform, we were able to train a neural network to classify the behavior of C.chilensis people. We use time series of accelerometers extracted from 5 monitoring campaigns of individuals of this species in their natural habitat. We designed a classification model using Keras to first distinguish motion from stillness. Subsequently, the objective is to distinguish different types of behavior: eating, walking, copulating, digging nests to lay eggs, etc. A concrete application of such a study is to find the eggs deposited by the female and to help in their conservation: by characterizing the signal of the accelerometer coming from the activity “digging nests to lay eggs” as well as the location of the individual from GPS, we can identify these deposition sites and protect the eggs from potential threats.
Left: an individual of C. chilensis monitored in its natural habitat with the device comprising inertial sensors. The device is placed on the hull and is secured with camouflaged tape to prevent predation. Right: raw data from x, y, z accelerometers (top), spectrogram (middle), frequency domain to extract spectral features (bottom). The three graphs on the right are taken from Edge Impulse.
We tested the classifier using a Wio terminal and Arduino code (see “Animal motion classifier using an NN trained with accelerometers” in the “Code” section) with the movement of a captive turtle. In this way, we can determine the result of the classification in real time and obtain the result (“Active” or “Inactive”) on our mobile phone via Bluetooth Due to the weight and dimensions of the Wio terminal, we will be using TinyML kits in the field.
Classifier using a Wio terminal and Arduino code detecting animal movement.
In field work, each individual is equipped with a device using camouflaged tape to avoid predation. Until now, the data was obtained once the memory of the device was removed. This procedure is the heaviest and longest task in the campaigns, since we have to locate animals that have spread over a relatively large natural area with rough vegetation. Above all, in doing so, we also strongly disturb the natural behavior that we want to study. Our goal is to use long-range, low-power communication protocols like LoRa to be able to see animal behavior in real time.
We can integrate SenseCAP K1100 sensors into both monitoring systems: the one developed by our team and the one implemented with TinyML kits. With enough LoRa devices, we could create a network connecting them. This would allow us to send the stored information without having to approach the individual closely and without having to touch the mounted equipment, which is carefully camouflaged and secured. LoRa also brings the possibility of scaling the study and monitoring several individuals or “nodes” in real time, and for example detecting possible collective behavior. It could also be used to update device settings, all these tasks non-intrusively from one or more monitoring/control centers.
Schematic connections of the Nano 33 BLE to implement a complete tracking device with long range communication. This allows building a network of monitored devices in real time.
The SenseCAP K1100 kit also opens new avenues for sensor fusion that we have not yet explored, and for addressing the question of the effect of temperature, humidity and light on turtle behavior. A better understanding of the movement of individuals is crucial to establishing guidelines that aid in their conservation. Taking advantage of the SenseCAP K1100, alongside animal tracking, we were able to measure environmental variables in the natural habitat in which they live. For example, one type of turtle shelter consists of burrows dug under the ground. Now we could measure the humidity of these places.
In addition to implementation in tortoises, we aim to use it in other animal species in the wild as long as the devices meet size and weight requirements. For example, we are starting a collaboration with groups that study the behavior of small lizards and small birds.
Finally, it is important to say that we really like this project and that there is still a lot to do. Contributing to the monitoring of different animal species using different technologies can be very valuable in learning more about the biology of each and establishing guidelines for their conservation in the case of vulnerable species.