Researchers Try Listening to Acquire Knowledge about Wildlife

In an effort to better understand how artificial intelligence (AI) might track animal life in recovering habitats, a group of scientists is listening to the sounds made by Ecuadorian forests. Scientists may use instruments like lidar and satellites to analyze broad swaths of land and assess the growth of new forests. It is more challenging to gauge the rate and volume of wildlife reintroduction in a given area. There are situations when an expert is needed to listen through sound recordings and identify animal sounds. Field expert on birds at the University of Wurzburg Biocenter in Germany is Jorg Muller. Is there another way, he wondered. He then looked to bioacoustics, which makes use of sound to discover more about the lives of animals and their habitats.

Scientists have been using the tool for a while now. More recently, however, it has been used by researchers to evaluate massive amounts of data more quickly when combined with computer learning. Muller and his group captured sounds of wildlife at locations in the Choco region of Ecuador. The settings they documented ranged from old-growth forests to places that had previously been utilized for cattle ranching and agriculture. First, they had professionals listen to the recordings and categorize the various animal sounds. After that, they measure the environment by looking at the sound quality. The calls that the program was trained on might be identified. Scientists questioned whether the computer could accurately determine how many distinct species of plants and animals there were in each area.

The team employed two separate controls to test if the program could accomplish that. One came from the specialists who listened to the audio recordings, and the other was predicated on illustrations of biodiversity that could be found in each habitat. Only 25% of the bird cries that experts could identify were recognized by the AI algorithm due to the restricted number of sounds available for training. However, the study found that it was still able to accurately quantify the levels of biodiversity in each area.