Tropical forests are noisy places, the air filled with the sounds of a wide diversity of birds and animals. They are also rather physically taxing thanks to daily thunderstorms, intense midday heat, and mountainous terrain. It makes it difficult for ecologists to perform manual field study projects, which are thus prone to a high failure rate. So research postgraduate Sarab Sethi – along with his supervisors Prof Rob Ewers, Dr Nick Jones, and Dr Lorenzo Picanali – have devised a real-time ecosystem monitoring device based around a Raspberry Pi.
“Our particular interest was in recording audio to capture the soundscape – or the combination of all the vocalising animals – as this is a rich data source that can be used to track birds, mammals, frogs, and more,” Sarab tells us. For this, the scientists required a device that could continuously record, compress, and upload huge amounts of data from the field while exploiting a patchy mobile signal to remotely transfer data to a server. “The Raspberry Pi was ideal as a low-cost, relatively low-power device with a usable amount of computing power and large support for a wide range of sensors,” says Sarab.
Jungle sounds
With their field site in Sabah, Borneo, in mind, they set about creating a system that could monitor the effects of oil palm plantations and logging on the region’s biodiversity by listening out for the sounds of animals. It involved using a Røde smartLav+ microphone to provide high-quality audio recordings, along with an external USB audio card, solar energy, and a 3G dongle to connect to the internet. The ultimate aim is to use artificial intelligence to pick up on the audio and make sense of the data.
For now, however, the Python-programmed software runs two threads concurrently: one continuously records data from a sensor and stores it in uncompressed files, and the second compresses this data and robustly uploads it using FTP to a remote server. “It is important that the device is networked to minimise the amount of times a scientist or research assistant has to go to visit the device to manually collect the data – freeing up time to be better spent on other more efficient and less exhausting tasks,” explains Sarab. “Large animals also love to play with (or more likely destroy) any equipment left in tropical forests. Continuously uploading data serves as an instant backup system.”
Testing times
The system is being tested at the Stability for Altered Forest Ecosystems (SAFE) project in Borneo where it’s been up-and-running since February 2018. It currently uses a network of twelve acoustic monitoring devices which are placed across a gradient of logging intensity, from old-growth untouched forest to oil palm plantations, allowing animals to be tracked in specific environments. “Our major headache has been the solar power system, as the quality and range of batteries available in this region of Malaysia is generally pretty poor and importing batteries across borders is difficult,” Sarab says.
“However, the monitoring device itself has fared surprisingly well, especially considering the near 100 percent humidity and regular movement due to shaking of trees, animal interference, etc. To date, we have collected over 15 000 hours of audio using these devices, and more is coming in each day.”
Quick facts
- The overall system costs £230 to make
- 64GB can store a month of animal sounds
- They’re exposed to temperatures of 2 to 31.5°C
- They withstand 614 mm of precipitation per month
- Powered by 20 to 30 W solar panels - attached to the top of trees to maximise sunlight
This project first appeared in The MagPi issue #76 and was written by David Crookes.
