A New Technology Can Remotely Analyze an Ecosystem’s Species By its Sound

By distributing networks of microphones to wetlands and forests around the world, biologists could track biodiversity in a whole new way

ARBIMON—a system of distributed recording stations and centralized analysis software—was used to track populations of the endangered plains coqui frog, in Puerto Rico. Photo by CoquiPR

Imagine you’re a scientist and you want to track the population of an endangered frog species in, say, the Puerto Rican rainforest.

In the old days, you’d have to write a proposal, win a grant, put together a team, trek out into the field and spend a few weeks or months manually collecting and cataloging samples. A few years later, if you wanted to know whether the frog population had recovered or gotten even smaller, you’d have to go through the same process all over again.

A new way of collecting this information, presented today by scientists from the University of Puerto Rico in the journal PeerJ, promises to make this process much easier, faster and more comprehensive. Their idea—a network of widely distributed microphones and web-based audio recognition software, which they call ARBIMON (for Automated Remote Biodiversity Monitoring Network)—could someday make it possible for us to eventually have real-time estimates on critical animal population levels in spots all over the world.

The researchers designed the distributed hardware part of the system to be built from relatively inexpensive, widely available components—such and iPods and car batteries—along with waterproof cases and solar panels, which would enable the microphones, once placed, to last several years. The idea is that a network of such microphones, with one placed roughly 50 square meters, could act as remote ears listening in on the ecosystem: Every ten minutes, each microphone will record one minute of the local ecosystem’s sounds (amounting to 144 recordings per day) and send it via a radio antenna to a nearby base station.

The system will be made from a network of widely placed recording stations
The system will be made from a network of widely placed recording stations and a unified interface that will allow biologists to access data. Image via ARBIMON

Each base station will then send the recordings on to a centralized server in Puerto Rico, from where they’ll be made public in near-real time at Arbimon.com. Simultaneously, software will analyze sounds from the recording to pick out the different noises made by different species. Using an existing bank of identified species calls, the software will assign particular sounds to particular birds, frogs and other creatures.

Verified users—perhaps a biologist working on research on a particular species, or a member of the general public with a background in birding, for example—can contribute to the project by listening to the recordings and verifying whether the software is correctly identifying sounds and matching them to right species. Over time, input by users will train the software to become more accurate.

Eventually, once the software is trained to identify each call, the researchers say it’ll be able to process more than 100,000 minute-long recordings in less than an hour. As a result, a biologist will be able access a constant stream of data on the levels of a specific species in spots around the world, or the fluctuating populations of various species in one ecosystem.

Initially, biologists can index certain frequencies of a species’ calls to known populations of that species in each location—for example, 400 coqui chirps per hour means that 10 coquis are in the area. Later on, when the frequency of calls changes, this data can be extrapolated to infer fluctuations in the population present.

In the published paper, the system’s capability was demonstrated by tracking populations of a number of birds, frog, insect and mammals species in Puerto Rico and Costa Rica over the past few years. At the Puerto Rico research site in the Sabana Seca wetland, the researchers focused on tracking populations of the Plains coqui frog, an endangered amphibian discovered in 2005 with a high-pitched, distinctive chirp.

Microphones were first installed there in 2008, and over the subsequent few years, the researchers trained the software to become increasingly accurate at analyzing the various sounds picked up and determining which were the plains coqui’s chirp. Eventually, scientists charted variations in the chirp’s frequency on both daily and seasonal timescales and were able to match these with surveyed data on changes in the coqui population.

One of the reasons these researchers are most excited about the new system is the way it’ll standardize and permanently store the audio samples indefinitely. 50 years from now, they say, if a conservation biologist wants to look back at the way populations of a species have fluctuated over time, he or she can simply access the recordings and have them analyzed. Not only will this help to track endangered populations, but could also pinpoint when invasive species began to dominate certain ecological niches.

The next step, according to the researchers, is installing these microphone setups in all sorts of ecosystems—every place where there’s a species that merits attention.

Get the latest Science stories in your inbox.