Fifty automated recording units scattered across the backcountry of Yellowstone National Park in the U.S. gather roughly 500,000 hours of ambient sound each year. Computer models then sift through those recordings for wolf vocalizations, reports Justin Catanoso for Mongabay.
Jeffrey Reed, a linguist and former software engineer, helped launch the Cry Wolf Project in 2023 within the long-running Yellowstone Wolf Project. It uses machine learning to pick out and analyze wolf calls. It turns the audio into spectrograms that reveal the structure of individual calls. So far, the project has isolated more than 8,000 continuous hours of wolf recordings, creating what researchers say is the largest database of wild wolf communication.
Tracking wolves has generally meant catching them and fitting them with radio collars. Dan Stahler, Yellowstone’s chief wildlife biologist, told Catanoso that doing so is expensive and carries risks for both wolves and people. Recorders can simply be left in the field. Researchers are testing whether differences in wolves’ voices can identify individuals and help track their movements.
The idea is also being tested in Europe. In northern Italy’s Po Valley, Osservatorio Lupi — “Wolf Observatory” — has deployed a dozen recording units near farms and towns. It hopes eventually to use the recordings to warn communities when wolves approach livestock. Researchers are exploring similar uses for brown bears and along wildlife migration corridors.
Reed has annotated roughly a fifth of the archive by hand to check the models’ results. For now, listening to wolves still involves a fair amount of human listening.
Banner image: Two wolf puppies from the Junction Butte Pack in Yellowstone National Park captured in a 2025 photo. Pups learn to howl early on by mimicking the vocalizations of their mothers. Image by Jeremy SunderRaj, National Park Service, courtesy of Jeff Reed.
