AI for Nature and Environment
The biodiversity crisis and climate crisis call for enormous ambition. They will require huge changes in society, such as land-use management. In order to monitor the natural world, and manage our relationship with it, we need rapid high-precision information and predictive modelling - AI can help to take the pulse of nature.
-
dr. Dan Stowell
Principal Investigator
We develop AI methods for monitoring wildlife and our environment:
- AI for animal sound recognition
- Develop AI methodology across a wide range of acoustic monitoring scenarios, for all our important species
- High-precision AI even for small-data bioacoustics issues (e.g., rare species, individuals)
- Animal vocal interaction - not only detecting animal calls, but also their patterns of interaction, and how these are affected by context
- Europe-wide AI nature monitoring (with Arise, and Horizon Europe funded projects)
- Ensure AI outputs can be integrated with later analysis e.g., statistical ecology
- Investigate AI for reducing carbon emissions (e.g., predictive monitoring in electricity generation)
- Implement environmentally responsible AI (e.g., low power/low CO2 emissions)
- Develop solutions that are widely applicable, e.g., useful in low-resource and poorer countries
Methods include deep learning, nonparametric Bayesian estimation (Gaussian processes), and other statistical machine learning methods, combined with acoustic signal processing techniques.
Projects
- Arise programme: Dutch nation-wide sound and image nature monitoring
- Horizon Europe funded projects (MAMBO, GUARDEN): Europe-wide sound and image nature monitoring
- Few-shot bioacoustic sound event detection
- Biodiversity monitoring in urban/construction sites (with Heijmans and others)
- Vocal interactions
Partners
This research line is developed as a close and ongoing collaboration with:
- Naturalis Biodiversity Centre
- Jheronimus Academy of Data Science