From chirps to songs, avian melodies can be as varied as a full orchestra.
However, ovenbirds — plump, terrestrial warblers — possess a very limited song repertoire.
The male ovenbird sings just one tune: a quick melody often characterized as teacher-teacher-teacher.
To most listeners, all ovenbirds appear identical in their singing, yet each has its own unique interpretation.
Currently, biologists are employing AI technology to monitor the health of ovenbird populations by analyzing their distinctive songs.
Research released this month in the journal Nature Methods utilized AI to distinguish individual ovenbirds in audio recordings obtained from Alberta and Pennsylvania.
Single-minded focus
The ovenbird’s “particular” song selection made it an ideal candidate for this study, explained lead researcher Sam Lapp, a postdoctoral researcher at the University of Colorado.
“They are quite consistent in singing one very specific song,” Lapp stated during an interview with CBC News.
“Many birds develop a vocal signature over their lifetime, often comprising various songs that form their repertoire.
“In contrast, the ovenbird has the advantage of maintaining one stable song over time, and it only sings one version of it.”
Ovenbirds perform their primary song during the spring season to establish their territory and attract mates, and this song remains a consistent part of their lives.
Under Lapp’s guidance while at the Department of Biological Sciences at the University of Pittsburgh, the research team gathered recordings from nearly 130 locations in Pennsylvania over a span of four years.
They trained an AI model to identify features that made songs distinctive and group similar ones together, based on the assumption that each bird returns to the same forest area each spring.
The team analyzed 9,072 hours of audio to monitor 405 individual birds. The model was then tested against another set of recordings collected from the mixed wood and boreal forests of Alberta during the 2016 spring breeding season.
Lapp noted that the model could associate songs with specific birds with an accuracy of up to 96 percent.
This method is significantly less labor-intensive than traditional leg banding, where scientists capture birds, carefully attach markers to their legs, and then attempt to recapture them the following spring.
He mentioned that the research initiative stemmed from a partnership with biologists in Alberta, who were working to locate birds across extensive areas using a grid of microphones.
“Astonishing” discovery
Lapp explained that human ears struggle to differentiate individual ovenbirds in real-time because their vocal differences occur too rapidly.
However, on spectrographs — visual displays of sound — the unique songs of each bird become clearer, Lapp noted.
“I began to wonder if my eyes could detect these patterns, could we teach a machine to recognize them and automate the process of identifying individuals?
“Even for those passionate about birdwatching and species identification, it’s surprising and exciting to realize this wealth of information has been overlooked… but is present if you listen closely enough.”
Erin Bayne, a biology professor at the University of Alberta and co-author of the study, has been closely monitoring ovenbirds for many years and is familiar with the challenges of annual population assessments.
During his doctoral studies, Bayne dedicated tens of thousands of hours to tracking banded birds, to the extent that he could recognize individuals by their distinct vocal differences.
“What that bird is saying over there differs slightly from the one over there,” he remarked. “I knew who was who.
“They are remarkably courteous in their singing. It’s quite amusing, actually. One bird will sing, and the other will patiently wait for it to finish before immediately singing afterward.”
Avian conversations
Bayne indicated that the study highlights the intricate nature of avian communication, revealing that exchanges are occurring even if we fail to grasp their language.
Similar to how humans possess unique speech patterns, biologists are discovering that various animals also cultivate their own vocal identities, according to Bayne.
“It’s not unexpected that individual birds or animals exhibit detectable differences. The more we explore this area, the clearer it will become that it is indeed true.”
While the model does have certain limitations, the researchers aspire to offer a less invasive method to assess the well-being of songbird populations simply by eavesdropping.
The AI system is not equipped to accurately monitor the varied calls of female ovenbirds or the basic vocalizations of young birds.
This current program would also struggle with tracking species such as magpies and mockingbirds that frequently mimic new sounds or have extensive song variations.
It is most effective for birds that maintain a consistent song throughout their lives — approximately 30 percent of songbird species — as well as specific species of bats and owls.
The researchers made the program code available on open-source platforms to assist other biologists in monitoring the survival of various songbird populations or even other animal species.
Richard Hedley, a statistical ecologist with the Alberta Biodiversity Monitoring Institute who was not part of the study, remarked on its significant potential.
Although AI acoustic monitoring has been employed for years to track birds at the species level, demonstrating its ability to differentiate between individual birds is a remarkable achievement, he noted.
There exists a substantial amount of existing acoustic recordings that could be utilized in innovative ways, he added.
“It can take countless hours just to capture a single bird. Thus, the idea that we can accomplish this by placing a microphone in the bird’s territory is highly promising.”