Solar-powered recording devices are eavesdropping on rainforests in Borneo to monitor biodiversity. The plan is to use聽artificial intelligence to聽automatically identify and record different animals and track changes over time.
Many current methods for measuring biodiversity are impractical, relying on humans to regularly change batteries and put out recorders, or expensive, requiring huge reams of data to be sent via satellites.
So at Imperial College London and his team developed a pared-down device built around a Raspberry Pi, a mass-produced cheap computer board. Solar panels peeking over the canopy provide power, while local mobile networks send data almost immediately to a remote server, accessible online. The devices can be fitted with multiple sensors to record multiple data sources, such as audio, photography and atmospheric information.
Advertisement
鈥淎utomatic monitoring is important if we are going to get a full picture of how species and ecosystems interact,鈥� says of the University of York.
Twelve monitors are currently nestled in the canopy picking up audio, in sections of rainforest with differing health levels. So far, over 10,000 hours of audio have been recorded and there are two plans of how best to analyse it.
One will work like a Shazam for species. Rather than identifying a song鈥檚 artist from a sound recording, artificial intelligence algorithms will recognise different animal calls and record how the populations shift over time.
The other takes a more holistic approach by creating an overall fingerprint of the forest based on the audio. By analysing how this alters as time passes, or in different areas of the forest, they can gauge the health of the ecosystem. Both methods will be used in tandem.
The work is part of a wider , documenting the effects of palm oil plantations on the forest. Since areas will be continued to be cleared, we want to find out which forest fragments will preserve the maximum amount of biodiversity, says Sethi.
Methods in Ecology and Evolution
