Intelligent Earth Seminar: Ella Browning (Oxford)

Leveraging machine learning for effective, large-scale passive acoustic monitoring of biodiversity and nature recovery

4 December 14:00
Earth Sciences: Lecture Theatre
thumbnail img

 

Ella Browning is a postdoctoral researcher in Ecoacoustics in the Leverhulme Centre for Nature Recovery at the University of Oxford, UK. She received her PhD in Environmental Science from University College London, UK and considers herself and interdisciplinary ecologist. Her research focusses on providing evidence for effective nature recovery measures, combining novel monitoring technologies, machine learning, and spatiotemporal statistical analysis. She has authored highly cited guidelines and reviews on the use of passive acoustics for ecological monitoring, collaborated closely on the development of machine learning algorithms for classifying bat echolocation calls or broader soundscape classification, and used deep learning to predict seabird dive locations. As a research scientist at the Bat Conservation Trust, a UK NGO, she developed cloud-based infrastructure for processing national bat passive acoustic survey data. In her current position, she is exploring the suitability of ecoacoustics to assess nature recovery measures across ecosystems and regions, globally, from above to below ground.

Leveraging machine learning for effective, large-scale passive acoustic monitoring of biodiversity and nature recovery

Effective nature recovery relies on evidenced knowledge of drivers of species population changes, and habitat health. Passive acoustic monitoring (PAM) facilitates rapid collection of ecological information at increasingly large scales relevant for assessing the success of nature recovery measures. This may focus on monitoring populations of taxonomic groups such as birds, frogs, or whales, (bioacoustics) or seek to quantify the health of the entire ecosystem based on the soundscape (ecoacoustics). In this talk, I will present some recent work using machine learning (ML) models to automate the classification of taxa specific vocalisations, such as bat echolocation calls, enabling national scale, or real-time biodiversity monitoring. I will focus on where ML approaches applied for training dataset generation or whole soundscape classification can address taxonomic and geographic biases which restrict PAM use in many regions. Finally, I will share our explorations into the mysterious world of soil ecoacoustics and how ML can help us decode belowground soundscapes to provide evidence of nature recovery.