The secret acoustic world of leopards: Advancing conservation through bioacoustics and machine learning

Professor Andrew Markham is co-author of a new study showing that leopards can be identified by their roars, opening the door to more work on how this vulnerable species communicates, and furthering conservation efforts. 

Leopards are solitary, nocturnal creatures that live across huge expanses of terrain, and gathering reliable data to observe the species is difficult. Studying leopards through the sounds they make – a technique known as bioacoustics and more typically used to monitor birds and marine species – has the advantage of allowing researchers to monitor much larger areas.  

The study was undertaken in a 450km2 area of Nyerere National Park in Tanzania, and researchers attached 50 camera traps paired with custom-built directional microphones to trees along roads and trails. Researchers were able to then match each leopard seen on the camera footage to the roar captured from the microphones. They extracted the key features of each leopard’s roar and used machine learning to analyse its temporal pattern, and found individual identification was possible with an overall accuracy of 93.1%. The study shows that using multiple forms of technology to record complementary data can exploit a wider variety of species traits than single technology studies alone. 

This was a fascinating collaborative project which exploited the strengths of each sensor type. Visual data is rich and allows precise individual identification from an animal’s spot pattern, but a camera’s field of view is limited to tens of metres. Acoustic data can be easily captured over hundreds of metres, but it was unclear whether it could serve as a useful tool for telling one leopard’s roar apart from another. This study leveraged both modalities to show that its voiceprints are indeed as unique as its spots. Professor Andrew Markham

Our success in using a combination of different types of technology could hopefully lead others to think about how to integrate different types of technology into their research, as the rich data this provides could really push science ahead and help us understand ecosystems and landscapes in a much more holistic way. Lead author Jonathan Growcott, University of Exeter 

The research was a collaborative effort between the University of Exeter, University of Oxford (Department of Computer Science and WildCRU), Lion Landscapes, Frankfurt Zoological Society, TAWIRI (Tanzania Wildlife Institute for Research) and TANAPA (Tanzania National Parks Authority). 

The secret acoustic world of leopards: A paired camera trap and bioacoustics survey facilitates the individual identification of leopards via their roars is published in Remote Sensing in Ecology and Conservation.