Leveraging AI for Sky Surveillance: Tracking Africa’s Biggest Bat Congregation

As evening descends, bats begin their feeding excursion from Kasanka National Park into the neighboring countryside. Image source: Christian Ziegler / Max Planck Institute of Animal Behavior

Artificial intelligence and computer vision have now contributed to the most accurate estimation yet.

Each year, a secluded forest in Zambia becomes the stage for an awe-inspiring natural spectacle. During November, straw-colored fruit bats embark on a migratory journey from various regions in Africa, ultimately uniting in a specific cluster of trees in Kasanka National Park. Mysteriously, these bats converge in a concentrated area of the park for a three-month stretch, leading to the creation of Africa’s largest bat colony.

Despite this impressive congregation, the actual bat count within this colony has always been uncertain, with estimations ranging anywhere from 1 to 10 million. The Max Planck Institute of Animal Behavior (MPI-AB), however, has recently introduced a new technique providing the most precise count to date. This method uses GoPro cameras to film the bats, and artificial intelligence (AI) is then deployed to recognize these creatures without the need for human intervention.

Published in the Ecosphere journal, this approach resulted in an estimated population of 750,000 to 1,000,000 bats in Kasanka, qualifying this gathering as the largest bat population by biomass globally.

“Combining cost-effective cameras with AI allows us to monitor large animal populations in ways previously unthinkable,” states Ben Koger, the study’s lead author. “This methodology will revolutionize our understanding of nature and how we strive to preserve it amidst rapid human development and climate change.”

Secret Gardeners of Africa

The straw-colored fruit bat is a standout even amid Africa’s diverse fauna. It’s considered by many as the most abundant mammal on the continent. Its annual journey of up to two thousand kilometers also makes it the most remarkable long-distance migratory flying fox. From an ecological standpoint, these attributes significantly matter. By spreading seeds over vast distances during flight, fruit bats play a critical role in reforesting degraded land—positioning them as a “keystone” species on the African continent.

Attempts to estimate the size of this crucial species’ colonies have been made, but the difficulty of manually counting such large populations has resulted in varied numbers. This has often been a source of frustration for Dina Dechmann, an MPI-AB biologist, who has dedicated over a decade to studying straw-colored fruit bats. Worried about a potential decline in fruit bat populations, Dechmann sought a tool that could reliably track changes in their numbers over time.

“Straw-colored fruit bats are Africa’s clandestine gardeners,” declares Dechmann. “Their seed dispersal connects the continent in a unique way. Losing this species would be catastrophic for the ecosystem. So, we need to know urgently if the population is diminishing.”

In pursuit of a solution, Dechmann connected with her long-term collaborators Roland Kays from NC State University and Teague O’Mara from Southeastern Louisiana University, along with Kasanka Trust, the Zambian conservation organization responsible for Kasanka National Park’s management and the protection of its bat colony. They contemplated whether advances in computer vision and artificial intelligence could enhance the accuracy and efficiency of large and intricate bat population counts. To explore this possibility, they reached out to Ben Koger, then a doctoral student at the MPI-AB, who specialized in using automated techniques for creating ecological datasets.

Reliable and Consistent Bat Census

Koger worked to devise a two-step method that could be utilized by researchers and conservation managers to efficiently quantify the complex system. Firstly, nine GoPro cameras were evenly installed around the colony to film bats departing from the roost at dusk. Secondly, Koger used deep learning models to automatically identify and count bats in the recorded videos. Testing the method’s accuracy, the team manually counted bats in a sample of clips and found that the AI achieved a 95% accuracy rate, even under dark conditions.

“The application of sophisticated technology for monitoring a colossal colony like Kasanka’s could be prohibitively expensive due to the extensive equipment needed,” explains Koger. “However, our study demonstrates that inexpensive cameras, combined with custom software algorithms, performed exceedingly well in detecting and counting bats at our research site. This finding is immensely important for future site monitoring.”

Over the course of five nights, the new method counted an average of approximately 750,000 to 1,000,000 bats per night. Although these results are lower than previous counts at Kasanka, the authors acknowledge that the study may not have captured the peak of bat migration, and some bats may have arrived after the counting period. Nonetheless, the research’s estimate signifies Kasanka’s colony as the world’s most massive bat congregation by weight.

“This is a breakthrough for counting and conserving large animal populations. We now have an efficient and repeatable method for monitoring animals over time. By using this identical method for annual animal census, we can definitively state whether the population is increasing or decreasing,” says Dechmann.

The Kasanka colony, which is facing threats from agriculture and constriction, urgently needs accurate monitoring, Dechmann states.

“Though it may seem that losing a few animals from a large population won’t make much difference, we need to maintain the population at substantial levels to preserve the ecosystem services they provide. The Kasanka colony isn’t just one among many; it’s a major colony that attracts bats from across the subcontinent. Losing this colony would be a tragedy for Africa as a whole.”

Reference: “An automated approach for counting groups of flying animals applied to one of the world’s largest bat colonies” by Benjamin Koger, Edward Hurme, Blair R. Costelloe, M. Teague O’Mara, Martin Wikelski, Roland Kays and Dina K. N. Dechmann, 29 June 2023, Ecosphere.
DOI: 10.1002/ecs2.4590

Frequently Asked Questions (FAQs) about Artificial Intelligence in Wildlife Conservation

More about Artificial Intelligence in Wildlife Conservation

• Max Planck Institute of Animal Behavior
• Straw-Colored Fruit Bat
• Kasanka National Park
• Application of AI in Wildlife Conservation
• Ecosphere Journal