© Ben Koger
How and why do groups of animals move and what role do certain individuals play in this? To study the behavior of community-dwelling species like zebra and co, researchers present a new method of data collection by drones and subsequent analysis by artificial intelligence. In this way, information about animals can be collected without interference and with comparatively little effort, which can benefit their protection and behavioral research, say the scientists.
Some steppe and scrubland dwellers, monkey species and many other animal species are known for their pronounced sociability: they form communities or larger herds that roam together, showing interesting patterns of group dynamics. How these systems work and the importance of certain factors is exciting from a biological point of view and can also be important for the protection of endangered species. However, research into this topic is difficult and time-consuming: certain aspects of group dynamics remain hidden or have to be laboriously recorded through evaluations. The collection of data by equipping individuals with data loggers and motion sensors is also time-consuming and also associated with stress on some of the endangered animal species.
Eagle Eyes and Computer Algorithms
For this reason, the scientists led by Benjamin Koger from the Max Planck Institute for Animal Behavior in Konstanz are working on an alternative. Now they are presenting their concept, which can be used in the case of animal species that live in open or sparsely vegetated landscapes: they use drones that look down on entire groups of animals from high above, collecting a range of data that can then be used evaluated in a special way to generate detailed information.
In concrete terms, this means that the researchers launch a drone equipped with computer-controlled recording technology and steer it into a position high above a group of animals so that it is not disturbed. The aircraft then records in detail the individual individuals, their movements and behavior as well as the three-dimensional features of the landscape including vegetation. After the drone returns, the data is then read out and analyzed by special computer algorithms that can automatically recognize information in recordings.
Comprehensive information on group dynamics
“We often recorded 20 or more different individuals at the same time. As a human, it would take weeks to determine where each and every individual is in a single half-hour video observation. So the first challenge was how to automatically recognize the animals we were interested in,” says Koger. This was made possible by artificial intelligence: The researchers used powerful deep learning algorithms that can "learn" to recognize individual characteristics of animals. Another challenge was the data cleansing for recording the movements of the animals, the researchers explain. Because the videos were also influenced by the drone movements and by distortions caused by landscape structures. However, the team apparently managed to overcome the hurdles and develop a practicable system, as shown by the example of Grevy's zebras in Kenya and Gelada monkeys in Ethiopia.
"The strength of our image-based method is that it represents a comprehensive solution," says Koger. Because the drone captures not only the animal groups, but also the landscape, a very comprehensive data set is obtained that contains information about the social and ecological context of all animals in the observed group. In this way, it can become clear without interference how social and spatial structures influence behavioral processes, such as decision-making and the exchange of information in groups.
In conclusion, senior author Blair Costelloe from the University of Konstanz says: “One of the strengths of our method is that it can be adapted to many different animal species and environments. I believe that this method can help us to develop an understanding of the mechanisms by which individual behaviors generate the overarching phenomena that are relevant to conservation," says the scientist.
Video: Illustration of the concept of the new process. © Ben Koger
Source: University of Konstanz, specialist article: Journal of Animal Ecology, doi: https://doi.org/10.1111/1365-2656.13904