How scientists are using AI to study blackbuck mating in the wild

In the grasslands of Tal Chhapar, Rajasthan, India, a herd of blackbucks, a species of antelopes, gather on the open ground under the scorching heat of the desert sun. Dark brown circles of varying sizes, made of dung, mark the ground. This is the territory marked by each male blackbuck on the mating ground called a lek. The male blackbucks swing their long winding horns back, stomp the ground and face off in short parallel walks. Those brave enough, clash their horns and engage in fierce fights kicking up dust around them. All this pomp and show is to win the hearts of the females watching from the periphery, in a unique mating ritual called lekking. Farther away, a team of scientists have an eagle’s perspective of the drama unfolding. They watch the show on their remote-controlled screens, with three drones capturing every movement of the blackbucks.

An eagle’s perspective of the blackbuck lekking ground showing the territories, captured by a hovering drone.

Researchers at the Max Planck Institute of Animal Behavior, and the Centre for the Advanced Study of Collective Behaviour, Konstanz, Germany, set up the MELA project, Mating Ecology of Lek-breeding Antelope to better understand lekking. Three postdoctoral researchers—Akanksha Rathore, Hemal Naik and Vivek Hari Sridhar—combine drones and artificial intelligence to track and study blackbucks like never before. Naik and Sridhar, shared insights on the groundbreaking technology driving their research, while their upcoming dataset paper reflects a great leap for interdisciplinary collaboration. 

Lekking is a rare and energy-intensive mating system, seen in less than two per cent of mammal species, with blackbucks being one of them. “Lekking is like speed dating”, says Naik. You can imagine a big room, which is the open lekking ground, with multiple tables, which are the different territories held by the males, and the females that check out each table until they decide on who the lucky one is – the catch here being that only a few males get lucky multiple times.

Need for Innovation

A machine-learning algorithm tracks and identifies individual blackbuck across the lekking ground.

Studying animal behaviour in the wild has always been challenging. Traditional tools like radio-telemetry may disrupt natural behaviours, stationary cameras have limited range, and manual observations can only focus on a few individuals at a time. A dynamic phenomenon like blackbuck lekking magnifies these limitations, where hundreds of individuals interact simultaneously on an open plain. Observing these interactions in a non-intrusive and comprehensive manner required an innovative solution. So, the scientists turned to drones and AI.

Naik says, “Drones are easy, simple to use and cost-effective” making it the best possible option at the time while also allowing for the collection of information-rich data. The researchers used three drones flying simultaneously at an altitude of 80m and captured the event on record in high resolution. They relay the drones to obtain continuous footage, which is later fed into machine-learning algorithms that track individual blackbucks frame-by-frame. Drones therefore allow the researchers to study all individuals simultaneously, capturing interactions across the entire lek.

Sridhar says,” One of the big advantages of using a tech like this is also that you can monitor all individuals simultaneously”. In manual observations, you can only focus on certain individuals in certain areas due to a human's limited mental bandwidth and attention. However, due to the nature and richness of this data, multiple people can work on different research questions simultaneously and over a longer time, thereby providing faster outcomes in the long run.

Ideal Case Study

Head raised and horns swept back, a male blackbuck performs a courtship display on the lek.

Photographed by Vivek H. Sridhar

Project MELA finds blackbuck lekking to be an excellent case study for developing and testing new tracking algorithms due to its unique features and consistent annual recurrence in roughly the same location. “Blackbucks are nice to study lekking by using this technology because they poop”, says Sridhar. Unlike many birds that lek using acoustic displays, blackbucks mark their territories with visible dung piles, which serve as reliable landmark features for identifying individual territories. Additionally, their lekking occurs in a large open space with a relatively stable population in a specific area, making them readily available for consistent observation throughout the season. These conditions are ideal for developing and training a tracking algorithm. Once the team perfects these algorithms, they could apply them to study other lekking species or animal behaviours in a non-intrusive way.

Until now, most animal tracking algorithms have been developed using publicly available videos and photos from platforms like YouTube. Naik and Sridhar say with the release of the dataset in December 2024, project MELA will mark a significant step forward in bridging the gap between biology and computer science. This dataset offers computer scientists a real-world challenge: tracking and reidentifying blackbucks across multiple drone footage. For biologists, it unlocks deeper insights into a rare and fascinating biological phenomenon. “It is the first and only dataset to study animal behaviour by tracking and reidentification using drones”, says Sridhar.

Challenges

Developing algorithms for wildlife tracking is not without hurdles. The algorithm currently struggles with reliably identifying individual animals, estimating their exact size, and distinguishing between similar-looking young males and adult females. “The best algorithm would be one that is trained on a larger dataset, but since a larger dataset does not yet exist, this marks an important step in that direction,” says Sridhar. Once a reliable algorithm for individually recognising animals, something comparable to facial recognition technology for humans, is developed, it could revolutionize not only the study of animal behaviour but also wildlife conservation and management.

A Bold Future

One of the most exciting aspects of this technology is its scalability. While the MELA project currently uses three drones, advanced algorithms could link footage across several drones, enabling researchers to monitor much larger areas. The limitations would then only be manpower and computational power. Blackbuck lekking served as the ideal testing ground for this tech, with implications extending far beyond; allowing scientists to track animal movements across vast landscapes, monitoring populations and aiding in conservation.

Interdisciplinary collaboration determines the success of this technology. Biologists offer behavioural expertise, while computer scientists refine algorithms to address real-world challenges. Project MELA hopes to inspire others to improve the technology, by releasing the dataset and expanding its capabilities and applications. Leveraging drones and AI opens doors to non-invasive, comprehensive studies of animal behaviour, potentially transforming wildlife field studies in Biology. It’s a bold vision, but one well within reach.