Soundscapes can provide a valuable insight into the health of a habitat. A Raspberry Pi can help conservationists listen in, explains Rosie Hattersley.
The louder the dawn chorus and the cacophony of wildlife, the better the locale and, by extension, life for the creatures that dwell there. This is a simplified take on the importance of acoustic monitoring, as conducted by Bugg.xyz using a Raspberry Pi-based system.
Bugg wildlife traps record the sounds of the forest, and are AI-trained to listen to the quietest creatures, as well as those that broadcast their whereabouts to all and sundry. In fact, it’s more of an acoustic census, creating detailed records of the particular soundscapes of contrasting habitats, noting changes that may indicate deforestation or a natural disaster leading to depleted biodiversity. Dr Sarab Sethi, an environment research scientist at Imperial College London, developed the Bugg acoustic monitoring system, which is now being used to record the sounds of forests from Norway to Taiwan, as well as Bali and Borneo.
Sound and vision
Sarab created his Bugg kit having read about field monitoring devices built using off-the-shelf components and based around a Raspberry Pi and which featured a microphone, a USB audio card, a plug-in internet dongle, and a weather sensor. Suitably inspired, Sarab set about designing “a continuous, long-term fully autonomous acoustic monitoring device” that could send details back to the lab for analysis and which was robust enough to be placed and remotely monitored in a variety of habitats. With price as well as the hardiness of the components to consider, Raspberry Pi was an obvious choice: “The bigger questions weren’t so much about whether a Raspberry Pi would work with it, but more around the field logistics of mounting one in a forest to withstand temperature swings, humidity swings, and long duration deployment.”
![Sarab [rear] and colleague install Bugg boxes in the Borneo rainforest, which is subject to man-made degradation that affects both wildlife and the forest itself](https://cyirc.org/wp-content/uploads/2023/05/bugg-xyz-acoustic-monitoring-for-conservationists_64622a1cf1ae4.png)
Helpfully, Sarab’s PhD supervisor had set up SAFE (Stability of Altered Forest Ecosystems), a rainforest degradation monitoring project in Borneo (acoustics.safeproject.net), which formed the ideal proving ground for the Bugg acoustic monitoring system.
“Whenever we’re setting up an ecological experiment, we’re looking at gradients,” says Sarab. The SAFE Project questions how habitat degradation in Bali and tropical forest scenarios affect biodiversity. Samples are taken from pristine virgin untouched forests to lightly logged and heavily logged forests and then ultimately, the oil part monocultures. “You can also look at machine learning embeddings of the audio or fingerprint, go and survey ten sites then figure out which soundscapes correspond to what types of biodiversity and what type of habitat quality.” By going to ten more sites, or visiting the same site on ten different days, Sarab can use Bugg to infer “just based on that minute of audio, what the habitat quality or the biodiversity or stability of that system is.”
In the event, Bugg more than held its own against the various moth-trapping and bug-detecting traps, and neatly complemented studies into how forest fragmentation and forest degradation affect biodiversity. The results were published in the journal Methods in Ecology and Evolution, plus a website that details how to go about building your own device. Unfortunately, acoustic monitoring sometimes reveals a lack of wildlife and, what appears from photos to be healthy rainforest, is actually an empty forest.
Anyone home?
The open-source Bugg kits are now used for multiple monitoring tasks. While vegetation sampling involves taking samples and measuring trees, Sarab’s Bugg monitor takes a more holistic approach, “using soundscapes as a whole to infer habitat health.” Once the microphone in one of the Bugg devices detects sounds, they are transmitted over a mobile network for machine learning-based identification and analysis by scientific researchers. Bugg is generally used to pick up the sounds of particular insects or bird species, so the Raspberry Pi 3-based kit can be trained to recognise and almost instantly identify what’s making each sound, be it migratory bird flocks over Arctic Norway or ants in an Indonesian rainforest. Some projects look at a particular habitat in great detail over time, as is the case in Borneo, with four distinct gradients of degraded forest, while in Norway there’s a focus on multiple sites with the aim to get a detailed overview of bird life and habitat choice across the whole country.
The MagPi #129 out NOW!
You can grab the brand-new issue right now from Tesco, Sainsbury’s, Asda, WHSmith, and other newsagents, including the Raspberry Pi Store in Cambridge. You can also get it via our app on Android or iOS.
You can also subscribe to the print version of The MagPi. Not only do we deliver it globally, but people who sign up to the twelve-month print subscription get a FREE Raspberry Pi Zero Pico W!
