Bat Sounds Offer Clues to Conservation and Climate Change
Bats might not have the iconic howl of the gray wolf or the movie-famous screech of a red-tailed hawk, but the calls they do make through echolocation generate diverse and unique sounds.
Now, scientists believe bat sounds could offer hints about how the winged mammals are dealing with climate change and help inform future conservation efforts.
That’s good news, because bats are in trouble. According to the International Union for Conservation of Nature, 26 bat species are critically endangered—meaning they face an imminent risk of extinction—51 others are listed as endangered, and 954 bat species are considered vulnerable. They’re also among the most under-studied of mammals.
To get a better grasp on identifying populations and different species in an environment, researchers are creating a reference library of bat noises to automatically identify species. The research, published in the journal Methods in Ecology and Evolution, is the first time automatic classification for bat calls has been attempted for a large variety of species.
“As bats use sound to find their way around, they leak sound out of themselves when they are flying, so we can eavesdrop on them to find out how many of them are there and how their populations are faring,” said study coauthor Kate Jones, a researcher at the Centre for Biodiversity and Environment at University College London. “To do that we need to understand what the bats sound like in Mexico, and we needed some way of recognizing them from each other.”
Jones said University College of London graduate student Veronica Zamora-Gutierrez, the lead author of the study, wanted to see how bats were adapting to climate change and land use changes but needed a good way of monitoring bat populations.
They focused their work in Mexico because the country contains high levels of biodiversity and a shockingly high number of threatened species—especially bats. The team used nets to catch bats, recorded their sounds, and then released them back into the wild. Over the course of a year, they gathered 907 recordings of 39 species of bats, and they obtained another 1,403 preexisting bat recordings from colleagues in Mexico.
Computer-based machine learning algorithms were used to compare the bat library they compiled with unknown bat calls; the system returned the closest match. The researchers’ system was able to accurately identify the sounds of a bat species 72 percent of the time and correctly pick out what bat family the sound came from 91.7 percent of the time.
“This is how human voice or face recognition software works,” Jones said. “In order for matches to be found, the algorithms must first learn how to recognize the known objects from each other.”
Machine learning algorithms have mostly been used to classify data at the level of species, but this is one of the first times they have been used to aid classification accuracy.
The researchers plan to return to Mexico, and use the tool to help keep tabs on how different bat populations are faring, and how they are responding to global climate change.
“Bats are important pollinators, seed dispersers, and regulators of insect populations, and bats are important members of ecosystems and are under threat,” Jones said.