The number of wild lions in Africa is estimated to have halved in the last 25 years with declines continuing. Population monitoring and the identification of priority areas for conservation are thus increasingly important.
Lions engage in roaring bouts to signal where they are and to convey their territorial boundaries. Research led by WildCRU’s Jonathan Growcott used machine learning to classify these bouts into four call types: moans, full-throated roars, grunts, and a newly identified intermediary roar. The full-throated roars are complex enough to be used to identify individual lions – a process that previously relied on expert human judgement. To address this challenge, the team created a tool to classify these roars automatically – achieving a 95% success rate when tested.
This work was published in 2025 and received considerable media coverage including by the BBC and National Geographic. The authors believe their tools will make it easier to monitor lion numbers, informing and supporting conservation efforts.
