summary: A new study reveals that African lions emit two types of roars, overturning long-held assumptions and opening the door to more precise wildlife monitoring. Researchers used machine learning to automatically distinguish between full-throat roars and newly identified intermediate roars with more than 95% accuracy, eliminating much of the human bias in speech identification.
This breakthrough enables reliable, non-invasive population tracking, greatly enhancing conservation efforts. As lion numbers continue to decline across Africa, AI-powered bioacoustics could become an important tool to protect vulnerable big cat populations.
Important facts:
- Two roar types: Lions emit both a full-throated roar and a newly identified intermediate roar.
- AI accuracy: Machine learning classified the type of roar with 95.4% accuracy, outperforming expert judgment.
- Conservation implications: Improved acoustic monitoring will support better population estimates and conservation strategies.
sauce: University of Exeter
A new study has found that African lions make not one, but two different sounds. This discovery will revolutionize wildlife monitoring and conservation efforts.
Researchers at the University of Exeter have identified a previously unclassified ‘intermediate roar’ alongside the famous full-throat roar.
This study ecology and evolutionused artificial intelligence to automatically differentiate between lion roars.
This new approach had an accuracy of 95.4%, significantly reducing human bias and improving identification of individual lions.
Lead author Jonathan Growcott from the University of Exeter said: “Lions’ roars are not only symbolic, but also unique characteristics that can be used to estimate population numbers and monitor individual animals. Until now, the identification of these roars has relied heavily on expert judgment, introducing potential human bias.”
“Our new approach using AI promises more accurate and less subjective monitoring, which is critical for conservationists working to protect dwindling lion populations.”
According to the International Union for Conservation of Nature’s Red List, lions are considered endangered. The total wild lion population in Africa is estimated at 20,000 to 25,000, but this number has halved in the past 25 years.
This study proves that lion’s roar bouts include both full-throat roars and newly named intermediate roars, challenging the long-held idea that there is only one type of roar.
These findings mirror similar advances in the study of other large carnivores, such as the spotted hyena, and highlight the growing potential of bioacoustics in ecological research.
Using advanced machine learning techniques, the researchers implemented this automated, data-driven approach to classify full-throat roars, improving their ability to differentiate between individual lions. The new process simplifies passive acoustic monitoring, making it more accessible and reliable compared to traditional methods such as camera traps and spore surveys.
Jonathan Growcott continued: “We believe that wildlife monitoring requires a paradigm shift and a major change in the use of passive acoustic technology. As bioacoustics improves, it will become essential for effective conservation of lions and other endangered species.”
The research is a collaboration between the University of Exeter, Oxford University Wildlife Conservation Department, Lion Landscapes, Frankfurt Zoological Society, TAWIRI (Tanzania Wildlife Research Institute), TANAPA (Tanzania National Parks Authority) and computer scientists from Exeter and Oxford.
Funding: This research was supported by the Lion Recovery Fund, WWF Germany, the Darwin Initiative and the UKRI AI Center for PhD Training in Environmental Intelligence.
Important facts:
answer: They identified a second distinct “intermediate roar” alongside the classic full-throat roar.
answer: Classifies roars with 95.4% accuracy, reduces human bias and improves individual identification.
answer: Accurate acoustic tracking helps estimate population size and increases conservation of rapidly declining lion populations.
Editorial note:
- This article was edited by the editors of Neuroscience News.
- Journal articles were reviewed in full text.
- Additional context added by staff.
About this AI and communication research news
author: Louise Vennels
sauce: University of Exeter
contact: Louise Vennels – University of Exeter
image: Image credited to Neuroscience News
Original research: Open access.
“Roar Data: Redefining the Lion’s Roar Using Machine Learning” J. Growcott et al. ecology and evolution
abstract
Roar Data: Using Machine Learning to Redefine a Lion’s Roar
To advertise territory and communicate within the pride, African lions emit a roar, and the symbolic roar is one of its components.
A lion’s full-throated roar has recently been proven to be a unique and distinguishing characteristic. At the same time, the frequency of large-scale passive acoustic monitoring surveys is increasing. Lion roars therefore have the potential to become a useful tool for counting individuals and estimating population density, to complement traditional survey techniques.
Currently, all-out scream selection relies heavily on expert reasoning and is therefore subject to human bias. We propose a data-driven approach to automatically classify a lion’s full-throat roar from the other sounds that make up its roar.
By using a two-state Gaussian Hidden Markov Model, we also showed that there are two types of lion roars (full-throat roars and newly named intermediate roars) that can be classified with 84.7% accuracy.
Furthermore, to describe lion calls, we use simple metrics: maximum frequency (Hz) and call length (seconds); K– Clustering was sufficient to classify lion roar types with high accuracy (95.4%), meaning that using data-based predicted full-throat roars improves the ability to identify individuals (F1 score 0.87 vs. manual full-throat roar classification 0.80).
Here, we establish an easy-to-understand implementation process that reduces knowledge gaps and makes passive acoustic monitoring more accessible in areas currently dominated by other monitoring technologies (e.g., camera surveys), opening up avenues for new research.
