Pig frog or howler monkey? Software IDs calls of the wild
A tool built by scientists at the University of Puerto Rico helped them spot signature croaking of the Eleutherodactylus juanariveroi, an endangered frog, in 37,000 out of 170,000 recordings.

Sometimes, the only way to spot a tiny critter like a small, very rare frog, or an insect the color of leaves, is by listening for its signature ballad to pierce the cacophony of forest sounds.

A team of scientists in Puerto Rico is developing a tool that will help scientists track and identify species using permanent audio recording stations and a Web-based tool that scours audio files for target species.

The group tested the Automated Remote Biodiversity Monitoring Network (ARBIMON) in locations in Costa Rica in Puerto Rico and found that they were able to identify nine different species just by the sounds they made.

"I'm thrilled with the results," Mitch Aide, a biologist at the University of Puerto Rico, who led the project, told NBC News.

Besides being a convenient cloud-based storage tool, its real strength is that it allows researchers to set up a search model tailored to a species. Software combs through hours and hours of the recordings for moments when their animal chirps in or croaks. When researchers want to track species over several years to estimate long-term impacts like climate chance or land use, this tool will be at their service.

Aide and team used ARBIMON to analyze five years of audio recordings collected at solar-powered permanent stations near the wetland reserve in Sabana Seca in Puerto Rico and La Selva Biological Station in Costa Rica. For the first time, Aide and a team report the efficacy of the system, in the July 16 edition of the online publication PeerJ, showing the tracking results of several species, including the mantled howler monkey, two varieties of insects, two birds (the great tinamou and the chestnut-mandibled toucan), and four amphibians

When this group tracked the pig frog, named after its characteristic vocalization, Aide wasn't expecting to hear many. "I've listened to hundreds of recordings and rarely came across it," he said. But when he let the software analyze 150,000 sound files, it found about 4,000 examples of pig frog calls. The software also picked up a recurring spike in frog calls every September, a trend that fewer, manually counted observations would have missed.

When the tools were tasked at finding another frog, Eleutherodactylus juanariveroi, more than 37,000 frog calls were weeded out of 173,526 recordings. "We could have never done that with an army of field biologists," Aide said.

The original recording station established in a wetland in Sabana Seca, Puerto Rico.
The original recording station established in a wetland in Sabana Seca, Puerto Rico.

"I like to think of this project as creating biodiversity weather stations," Aide said, since they're collecting data day in and day out for days and years. When the recordings are uploaded on the website, anyone can listen to the audio files recorded in wildlife locations around the world. "You're creating the equivalent of a museum piece," he said.

Aide, who is a tropical plant biologist, got into the audio software side of things he says, because the community lacks long-term data about animal movement and behavior, to see how climate change and land use are affecting fauna. Data is aplenty, but there wasn't a good way to look at all of it at once. "There are lots of people wandering around with external hard drives [that] have no good way of managing their recordings," Aide said.

The website currently hosts 1.3 million sound files, about a minute long. Uploads are coming in from two permanent stations still operational in Puerto Rico, and portable recorders that biologists are carrying around in sites in Brazil, Argentina, Costa Rica and in Puerto Rico.

The ARBIMON website allows registered members to upload unlimited audio files, and are charged 1 cent per minute. For anyone to stop by and listen in, it's free.

The authors of "Real-time bioacoustics monitoring and automated species identification" in PeerJ include Carlos Corrada-Bravo, Marconi Campos-Cerqueira, Carlos Milan, Giovany Vega and Rafael Alvarez, in addition to Mitch Aide.

Nidhi Subbaraman writes about science and technology. Follow her on Facebook, Twitter and Google+.